BINARY TERNARY QUAD TREE PARTITIONING FOR JVET
1. An apparatus for coding video data comprising one or more processors configured to:
- receive a bitstream indicating how a coding tree unit was partitioned into coding units according to a partitioning structure that allows root nodes to be partitioned with at least one of quadtree partitioning, binary partitioning, and/or ternary partitioning;
parse said bitstream to determine how the coding tree unit was partitioned using the quadtree partitioning, binary partitioning, and/or ternary partitioning;
parse said bitstream to determine a symmetry of said partitioning, to determine if at least one of asymmetric partitioning and/or symmetric partitioning was used to partition a parent coding unit into child nodes, wherein symmetric partitioning splits a parent node in to child nodes of equal size, and asymmetric partitioning splits a parent node in to child nodes of unequal size;
identify each of the child nodes within each respective parent unit, wherein a node can be recursively partitioned in to smaller nodes,wherein both ternary and binary partitioning are allowed in the partitioning structure in either order during a recursive partitioning of a root node or parent node; and
decode the identified coding units using JVET.
A method of partitioning in video coding for JVET, comprising representing a JVET coding tree unit as a root node in a quadtree plus binary tree (QTBT) structure that can have quadtree, ternary, or binary partitioning of the root node and quadtree, ternary, or binary trees branching from each of the leaf nodes. The partitioning at any depth can use asymmetric binary partitioning to split a child node represented by a leaf node into two or three child nodes of unequal size, representing the child nodes as leaf nodes in a binary tree branching from the parent leaf node and coding the child nodes represented by final leaf nodes of the binary tree with JVET, wherein further partitioning of child nodes split from leaf nodes via asymmetric binary partitioning may be restricted depending on the partitioning type of the parent node.
- 1. An apparatus for coding video data comprising one or more processors configured to:
receive a bitstream indicating how a coding tree unit was partitioned into coding units according to a partitioning structure that allows root nodes to be partitioned with at least one of quadtree partitioning, binary partitioning, and/or ternary partitioning; parse said bitstream to determine how the coding tree unit was partitioned using the quadtree partitioning, binary partitioning, and/or ternary partitioning; parse said bitstream to determine a symmetry of said partitioning, to determine if at least one of asymmetric partitioning and/or symmetric partitioning was used to partition a parent coding unit into child nodes, wherein symmetric partitioning splits a parent node in to child nodes of equal size, and asymmetric partitioning splits a parent node in to child nodes of unequal size; identify each of the child nodes within each respective parent unit, wherein a node can be recursively partitioned in to smaller nodes, wherein both ternary and binary partitioning are allowed in the partitioning structure in either order during a recursive partitioning of a root node or parent node; and decode the identified coding units using JVET.
- View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 10)
- 11. A method of decoding a JVET bitstream, the method comprising:
receiving a bitstream indicating how a coding tree unit was partitioned into coding units according to a partitioning structure that allows root nodes to be partitioned with at least one of quadtree partitioning, binary partitioning, and/or ternary partitioning; parsing said bitstream to determine how the coding tree unit was partitioned using the quadtree partitioning, binary partitioning, and/or ternary partitioning; parsing said bitstream to determine a symmetry of said partitioning, to determine if at least one of asymmetric partitioning and/or symmetric partitioning was used to partition a parent coding unit into child nodes, wherein symmetric partitioning splits a parent node in to child nodes of equal size, and asymmetric partitioning splits a parent node in to child nodes of unequal size; identifying each of the child nodes within each respective parent unit, wherein a node can be recursively partitioned in to smaller nodes, decoding the identified coding units using JVET.
- 12. A method of multiple partitioning of a video coding block for JVET, the method comprising:
representing a JVET coding tree unit as a root node in a binary, ternary, and quadtree (BTQT) structure having a quadtree branching from the root node and any of quadtree, binary or ternary trees branching from each of the quadtree'"'"'s leaf nodes; using asymmetric partitioning to split a child node represented by a quadtree leaf node into child nodes of unequal size; using symmetric partitioning to split a child node represented by a quadtree leaf node into child nodes of equal size, wherein the asymmetric partitioning and the symmetric partitioning occur in any order and within the same BTQT structure; and representing the child nodes as leaf nodes in each quadtree, binary, and/or or ternary tree branching from the quadtree leaf node.
This Application claims priority under 35 U.S.C. § 119(e) from earlier filed U.S. Provisional Application Ser. No. 62/341,325, filed May 25, 2016, U.S. Provisional Application Ser. No. 62/408,696, filed Oct. 14, 2016 and from earlier filed U.S. Provisional Application Ser. No. 62/414,060, filed Oct. 28, 2016, both of which are hereby incorporated by reference.
The present disclosure relates to the field of video coding, particularly a block partitioning scheme for JVET that allows binary, ternary, and quadtree partitioning, in symmetric and/or asymmetric manners, in a binary, ternary, quadtree (BTQT) structure.
The technical improvements in evolving video coding standards illustrate the trend of increasing coding efficiency to enable higher bit-rates, higher resolutions, and better video quality. The Joint Video Exploration Team is developing a new video coding scheme referred to as JVET. Similar to other video coding schemes like HEVC (High Efficiency Video Coding), JVET is a block-based hybrid spatial and temporal predictive coding scheme. However, relative to HEVC, JVET includes many modifications to bitstream structure, syntax, constraints, and mapping for the generation of decoded pictures. JVET has been implemented in Joint Exploration Model (JEM) encoders and decoders.
The present disclosure provides a method of partitioning a video coding block for JVET, the method comprising receiving a bitstream indicating how a coding tree unit was partitioned into coding units according to a partitioning structure that allows root nodes to be partitioned with at least one of quadtree partitioning, binary partitioning, and/or ternary partitioning, parsing said bitstream to determine how the coding tree unit was partitioned using the quadtree partitioning, binary partitioning, and/or ternary partitioning, parsing said bitstream to determine a symmetry of said partitioning, to determine if at least one of asymmetric partitioning and/or symmetric partitioning splits a parent node into child nodes, wherein symmetric partitioning splits a parent node in to child nodes of equal size, and asymmetric partitioning splits a parent node in to child nodes of unequal size, and identifying each of the child nodes within each respective parent unit, wherein a node can be recursively partitioned in to smaller nodes. Both ternary and binary partitioning are allowed in the partitioning structure in either order during a recursive partitioning of a root node or parent node. The method also includes decoding the identified coding units using JVET.
The present disclosure also provides an apparatus for coding video data comprising one or more processors configured to perform the techniques disclosed herein.
The present disclosure also provides a method of partitioning a video coding block for JVET, the method comprising representing a JVET coding tree unit as a root node in a binary, ternary, and quadtree (BTQT) structure having a quadtree branching from the root node and any of quadtree, binary or ternary trees branching from each of the quadtree'"'"'s leaf nodes, splitting a child node or block represented by a quadtree leaf node into child nodes of unequal size, and using symmetric partitioning to split a child node represented by a quadtree leaf node into child nodes of equal size. The asymmetric partitioning and the symmetric partitioning may occur in any order and within the same BTQT structure. The method includes representing the child nodes as leaf nodes in each quadtree, binary, and/or or ternary tree branching from the quadtree leaf node.
Further details of the present invention are explained with the help of the attached drawings in which:
Disclosed herein is a flexible coding block structure for JVET that allows multiple tree structures called binary, ternary, quad tree partitioning (BTQT) structures. As disclosed, embodiments for BTQT support asymmetrical binary partitioning (AP). In embodiments, BTQT is designed such that it does not have many restrictions and does not require excessive complexity.
Frames of a video sequence, or more specifically the coding tree units within each frame, can be encoded and decoded using JVET. JVET is a video coding scheme being developed by the Joint Video Exploration Team. Versions of JVET have been implemented in JEM (Joint Exploration Model) encoders and decoders. Similar to other video coding schemes like HEVC (High Efficiency Video Coding), JVET is a block-based hybrid spatial and temporal predictive coding scheme. During coding with JVET, a frame is first divided into square blocks called CTUs 100, as shown in
Generally, the decision whether to code a picture area using inter-picture (temporal) or intra-picture (spatial) prediction may be made at the coding unit level, so a slice is divided into a sequence of CTUs during the encoding process. Each CTU 100 in a frame can be partitioned into one or more CUs (Coding Units) 102. CUs 102 can be used for prediction and transform as described below. Unlike HEVC, in one or more embodiments in JVET the CUs 102 can be rectangular or square. Further, in one or more embodiments, JVET CUs can be coded without prediction units, or transform units, which removes the separation of CU, PU, and TU concepts and supports flexibility for additional CU partition shapes to better match the local characteristics of video data. In one or more embodiments, the encoder includes a process to try a plurality of coding modes that do not require further partitioning for prediction and transform, which can reducing coding costs, complexity, and time. In some cases, the partitioning of the CUs 102 may still be referred to as a partitioning in to prediction units or transform units. The CUs 102 can be as large as their root CTUs 100, or be smaller subdivisions of a root CTU 100 as small as 4×4 units.
In JVET, a CTU 100 can be partitioned into CUs 102 according to a quadtree plus binary tree (QTBT) scheme in which the CTU 100 can be split into square units according to a quadtree or binary partitioning, and those square units can then be split horizontally and/or vertically according to quadtree or binary trees. For example, child nodes of a binary split may be further split using binary partitioning. Thus, WET'"'"'s QTBT coding scheme may be more flexible than HEVC'"'"'s quadtree structure (QT) because in addition to QT partitioning from a parent node in to equal sized children nodes, QTBT enables binary partitioning (BP) from a parent node into children nodes horizontally or vertically. The QTBT structure introduced, therefore, represents a tree in which a parent node can split using quadtree into four child nodes, or by using other partitioning methods as described in more detail below. Each of the child nodes may become a parent node for another split into additional child nodes. Though, in one or more embodiments, once a binary split occurs, further quadtree partitioning may not be allowed.
In some embodiments JVET can limit binary partitioning in the binary tree portion of a QTBT to symmetric partitioning, in which parent nodes can be divided in half either vertically or horizontally along a midline. By way of a non-limiting example,
It is noted that the term coding block is generally used to define a region covering a particular component (e.g. luma), and may differ in spatial location when considering chroma sub-sampling such as 4:2:0. A block is also used to describe the shape of CTUs and CUs, but the terms should be clear from the context. For example, as described in more detail below, a CU sometimes includes coding blocks (CBs) of different colour components, e.g. one CU contains one luma CB and two chroma CBs in the case of P and B slices of the 4:2:0 chroma format and sometimes consists of a CB of a single component, e.g., one CU contains only one luma CB or just two chroma CBs in the case of I slices. In one or more embodiments, the CTU includes a luma coding tree block and corresponding chroma coding tree blocks and syntax elements, which can be partitioned in to smaller blocks using a tree structure and quadtree-plus signaling, where the tree structure syntax of the CTU specifies the size and positions of its luma and chroma coding blocks.
A CTU, such as CTU 100, is a type of coding tree block that represents an entire N×N square of picture samples for some value of N, where the first root node at the highest hierarchy of the tree structure (depth 0) is associated with the entire coding tree block or CTU. A root node with depth 0 may be partitioned in to two or more child nodes, also referred to herein as leaf nodes. For example, a quadtree node could be further partitioned by a binary tree, where the quadtree leaf node is also the root node for the binary tree having binary tree depth as 0. Further, while each of the nodes split from a prior node may be child nodes or leaf nodes, such nodes are also parent nodes when the child node is further partitioned into additional child nodes. Thus, a node may exist as both a parent node and a child node, and a root node may also be referred to herein as a parent node.
The tree structure is split among multiple branches and a single branch may include several nodes, until a final leaf along each branch is reached, which is referred to as the coding unit. Thus, a final leaf node in a recursively split branch is also referred to herein as a terminating leaf node or final child node. Collectively these final leaf nodes represent the final set of coding units (CUs) that make up the CTU, and will not be further partitioned. Thus, where a leaf node represents a coding unit the terms may be used interchangeably, and a final leaf node should be understood to correspond to the final coding unit along the respective branch.
The root node in
In one or more embodiments, NET can allow either symmetric binary partitioning or asymmetric binary partitioning in a binary tree portion of a QTBT. In one or more embodiments, symmetric or asymmetric partitioning may occur at any depth in the tree structure. Asymmetrical motion partitioning (AMP) was allowed in a different context in HEVC when partitioning prediction units (PUs). However, for partitioning CUs 102 in NET according to a QTBT structure, asymmetric binary partitioning can lead to improved partitioning relative to symmetric binary partitioning when correlated areas of a CU 102 are not positioned on either side of a midline running through the center of the CU 102. By way of a non-limiting example, when a CU 102 depicts one object proximate to the CU'"'"'s center and another object at the side of the CU 102, the CU 102 can be asymmetrically partitioned to put each object in separate smaller CUs 102 of different sizes.
Syntax can be coded in the bitstream that indicates how a CTU 100 was partitioned into CUs 102. A bitstream may comprise a sequence of bits, in the form of a network abstraction layer (NAL) unit stream or a byte stream that forms the representation of coded pictures and associated data forming one or more coded video sequences. A syntax may refer to an element of data represented in the bitstream, also referred to as a syntax element. Each syntax structure including zero or more syntax elements present together in the bitstream may be included in the NAL unit, and they may be presented in a specified order.
As described in more detail below, a syntax element may be a flag that is a variable or single-bit syntax element that can take one of two possible values: 0 and 1, or the syntax element may have multiple descriptors and or values available. A syntax may be a statement or element with an associated descriptor. The syntax element may be an expression used to specify conditions for the existence, type, and quantity for the syntax. When a syntax element appears, it may specify which syntax element to parse from the bitstream. A bitstream pointer may be advanced to the next position beyond the syntax element in the bitstream parsing process.
By way of a non-limiting example, syntax can be coded in the bitstream that indicates which nodes were split with quadtree partitioning, which were split with symmetric binary partitioning, and which were split with asymmetric binary partitioning. Similarly, syntax can be coded in the bitstream for nodes split with asymmetric binary partitioning that indicates which type of asymmetric binary partitioning was used, such as one of the four types shown in
The example child nodes and partitioning thereof included herein demonstrate that a child node may be split, or partitioned, into smaller, final coding units. A root node may represent the entire CTU, and is also referred to as a parent node when partitioned into child nodes. Thus, a node may represent both a parent block and a child block because the child nodes become parent nodes if/when they are partitioned.
While additional embodiments are depicted in this disclosure such as described below with respect to
As depicted in
In embodiments where the use of asymmetric partitioning is limited to splitting quadtree leaf nodes, search complexity may be reduced and/or overhead bits limited. In one or more JVET embodiments, a condition may be imposed that prohibits quadtree partitioning in a binary tree node such that a binary partitioned child node cannot be further partitioned using quadtree. In either scenario, asymmetric partitioning in a QTBT structure may be employed using the disclosed techniques to limit overhead signaling, such that excessive overhead signaling is not required. For example, in embodiments in which only quadtree leaf nodes are split with asymmetric partitioning, the use of asymmetric partitioning can directly indicate the end of a branch of the QT portion without other syntax or further signaling. This type of signaling may also be referred to as implicit signaling, i.e., to signal what option is selected by using available information at the decoder to make a decision without an explicit signal or flag. Such technique may be more efficient when another syntax element inherently includes information that the explicit flag would redundantly provide. Similarly, for embodiments in which asymmetrically partitioned nodes cannot be split further, the use of asymmetric partitioning on a node can also directly indicate that its asymmetrically partitioned child nodes are final CUs 102 without other syntax or further signaling.
One or more embodiments disclosed herein include schemes where asymmetrical partitioning is only allowed at the bottom level of a quadtree portion of a proposed QTBT structure, allowing no further splits for asymmetrical partitioned nodes. Such limitations, e.g., no further partitioning allowed to a child node if its parent was asymmetrical partitioned, may be imposed based on encoder/decoder complexity. Further, as described in more detail below, for the dimensions of a child block that is a product of asymmetric partitioning and is not a square, further asymmetric splits that create non-integer ratios with respect to the other child nodes and the root parent node may require additional coding complexity to design a transform process for the non-integer size. Odd size transform is not supported.
If a node uses quadtree partitioning, therefore, the node may have children and the children can use quadtree partitioning, binary partitioning, and/or asymmetric partitioning. But, in embodiments where asymmetrical partitioning is only allowed at the bottom level, if a node uses asymmetric partitioning then no further splits are allowed for that node. Further, in embodiments, if a node uses binary partitioning, the node may have children but its children nodes can only use binary partitioning.
In alternate embodiments, such as when limiting search complexity and/or limiting the number of overhead bits is less of a concern, asymmetric partitioning can be used to split nodes generated with quadtree partitioning, symmetric binary partitioning, and/or asymmetric binary partitioning. In some instances, a flexible approach to partitioning may be too flexible and require highly complex encoder operations, or they may be too restrictive and result in suboptimal coding performance.
As described above, the partitioning employed may be subject to certain restrictions. For example, a node may be recursively partitioned until its level or level depth reaches a permissible maximum level or level depth. The splitting process for the node may be recursively indicated, and syntax elements for the last coding unit that is not further partitioned, or a final leaf node, in the level may be defined. Recursive splitting refers to the successive partitioning that occurs within a single node. When a child node is split it becomes a parent node having child nodes, and each successive child node that results from successive partitioning is a result of recursive partitioning.
As disclosed herein, flexible coding structures that allow asymmetrical partitioning as part of a BTQT structure may be modified. Techniques disclosed and further described below include an example of partitioning using quadtree, then partitioning the child units resulting from the quadtree partition using binary or ternary partitioning.
Block structures are critical for video compression systems. A partitioning structure that supports binary tree, ternary tree, and quad tree may provide a significant coding gain over a QTBT structure, however overhead bits may be too high due to the amount of redundancy in arriving at certain partitioning outcomes.
As illustrated by the example embodiments, partitioning may be restricted in certain ways and such restrictions may be used in combination. The restrictions may be imposed to use the partitioning techniques in a more efficient by imposing rules on the partitioning. For example, a ternary split may not be allowed to occur as a result of successive asymmetric binary splits. As further described below, a specific syntax for ternary split may be included and may avoid the need for successive asymmetric binary splits to achieve a ternary split.
For example, an example of partitioning for a proposed BTQT may include the following rules:
- 1. if a node uses quadtree partitioning, the child nodes may be square nodes that can use at least one of quadtree partitioning, binary partitioning, ternary partitioning and/or asymmetric partitioning;
- 2. if a node uses asymmetric partitioning (such as the 3:1 asymmetric partitioning shown in
FIG. 5D), the child nodes may include a binary node and a restricted node, where the child binary node can further use binary partitioning, ternary partitioning, or asymmetric partitioning, and
- 3. if a node uses binary partitioning or ternary partitioning, the child nodes are binary nodes and they can each use binary partitioning, ternary partitioning, or asymmetric partitioning.
In embodiments, the restricted node (the larger child node) in ternary partitioning may disallow further partitioning of the restricted node. In alternate embodiments, further partitioning of the restricted node is allowed, or it is allowed in some circumstances and not in others. There are multiple possible rules for restricted nodes. For example, binary partitioning, ternary partitioning, or asymmetric partitioning may be applied to a restricted node when such partitioning is allowed. In embodiments, only ternary partitioning is allowed for the restricted node.
The partitions illustrated by
To add a complexity control that supports modular implementation, a parameter called size discrepancy limit (SDL) may be employed to limit smallest size of BN and RN. This parameter sets a minimum size of child node based on ratio of its horizontal and vertical dimensions, referred to as size discrepancy (SD), as described in equation 1 below. For example, in example embodiments when SDL is set to 4, a BN of size 32×4 or a RN of size 12×64 is not allowed. When a SDL parameter is enforced, a max BT depth parameter can be dropped in one or more embodiments.
It may be desirable to maintain dimension having integer values. For example, if a parent node dimension is 3N, the children should have dimensions of 2N and N. In another example, if a parent node dimension is 4N, its children can have dimensions of 3N and N or 2N and 2N, as long as the N multiplier is an integer value. Thus, in one or more embodiments, the conditions of partitioning may be defined to prevent non-integer child parts of the whole where the smallest coding unit is assigned a value of 1, where the whole is a square parent coding unit. In other words, conditions may be set such that a ratio of child unit sizes that together make up a parent square node will have a ratio equal to: 1(size of the smallest coding unit):(an integer value). Thus, the child nodes that result from partitioning will have an integer-sized value in a ratio of its size to the smallest child unit having size 1 part of the whole of any square parent coding unit in the child node'"'"'s ancestry.
Using the embodiment in
Starting with the whole parent square node A, first a 3:1 partition occurs resulting in a larger child node B (represented by 102j+102p+102q) and smaller child node C (represented by 102r+102s). Smaller child node C is further 3:1 asymmetrically partitioned into child units D (102r) and E (102s). The larger node B is partitioned using 1:2 asymmetrical partitioning, into larger child unit F (102j) and smaller child node G (102p+102q). Smaller child node G is symmetrically binary partitioned in to H 102p and I 102q.
In accordance with the example rules above and if there is a requirement for an integer ratio, each time a child unit is partitioned, the resulting child units within a parent square node must have a size that has an integer ratio value relative to the size of the smallest coding unit assigned a value of 1 and relative to the size of the parent square unit. For example, assume parent square node A has 4 parts and is partitioned according to a ratio of 3N:1N, child node B (102j+102p+102q) represents 3 parts of square node A (3N) and child node C represents 1 part of square node A (1N), respectively, for a ratio of 3:1. When child node B having size 3N is 2:1 partitioned in to F and G, F (102j) represents 2 parts of square node A (1N) and G (102p+102q) represents 1 part of square node A (1N), respectively, for a ratio between nodes C:F:G of 1 where child node C represents 1 part of the whole for a total of 4. Thus, assigning the smallest coding unit in the plurality of coding units that make up a parent square node to a value of 1 part of the whole, each child node has an integer size representation relative to the smallest coding unit and relative to the whole. Further, the sum of the assigned ratios for the plurality of coding units adds up to the whole.
In accordance with the rules, at the point of partitioning child node B (represented by 102j+102p+102q) could not be binary partitioned, as B has size 3N of the whole having size 4N in a ratio of 3:1 with unit C. A binary partition of B would result in child nodes each having a size 1.5 of square node A relative to C'"'"'s size of 1. Thus, the non-integer size child units relative to the parent square coding unit is disallowed according to these rules. Limits on a ratio with non-integers such as the example 1.5:1.5:1 above acknowledge encoding and decoding processes that may rely on non-integer ratios, thus avoiding further modification to support non-integer ratios. Further limits on partitioning may be in place to meet the needs of encoders in terms of memory and computational requirements.
In another example, child unit G includes child units 102p and 102q that result from a binary partitioning of portion G. Assuming portions C and F make up the rest of parent square node A, each child unit H (102p) and I (102q) represent the smallest unit, each assigned a size 1 for purposes of the ratio. Thus the ratio of H:I:C:F=1:1:2:1, all integer values which, when summed, equal the 4 parts of the whole square node A.
In order to maintain integer ratios of a child coding block sizes relative to a root (or parent) node with an N×N dimension, the partitioning of a node having a size that is an odd value relative to parts of the whole is limited, such that the node may be a restricted node as described above. On the other hand, binary partitioning of a root square node would result in two child nodes each having 1 part of 2 of the root square node for a ratio of 1:1. If one of those child nodes is further partitioned using 3:1 partitioning, the smallest child unit in the root square node assigned a size of 1N, the ratio of parts is 4:3:1 for a total of 8=2N. Thus, the root square node can be deemed to have 8 parts, the larger child node having 3 parts of 8, the smaller node from the 3:1 partitioning having 1 part of 8, and the child node resulting from the first partition having 4 parts of the 8. Thus, even with limitations in place, there are many flexible options for both asymmetric and symmetric partitioning within a CTU.
In one or more embodiments, non-integer ratios in contrast to the embodiments above is contemplated, however it is acknowledged that modifications to the existing HEVC encoder and decoder would be required for the coding units of such varying sizes.
A coding tree unit may be represented and coded as recursive tree structure, and the tree structure may be expressed by a series of flags. For a BTQT structure, the first syntax to be encoded may be a flag indicating whether there will be further partitioning or not. If there is further partitioning, a second syntax may be used to indicate whether to use horizontal or vertical partitioning. Then a third syntax encoded may be a flag indicating whether symmetric partitioning or asymmetric partitioning is to be used; e.g., symmetric, BP, asymmetric binary partitioning shown in
If “No QP split” is signaled by the flag.quadtree_partitioning syntax element, quadtree partitioning is not used. The flag.partitioning_method flag may be a second flag that, once determined by a first flag that QP is not used, the second flag indicating which non-QP method will be used or not. Alternately, the flag.partitioning_method flag may be used to indicate both whether the quadtree partitioning is used or not (e.g., value 0 for quadtree and any other value for “non-QP”) and also signal the type of non-quadtree partitioning method used (e.g., value 1-4, 1=binary, 2=AP, 3=ternary, etc) is used.
As shown in
If there is further partitioning using a non-QP partitioning method, a third syntax may be signaled to indicate whether to partition in a horizontal or vertical direction. Both examples of a vertical partitioning direction and a horizontal partitioning direction are provided in
For either of the identified vertical or horizontal partitioning directions, an additional syntax may be used to identify which partitioning method is applied, e.g., asymmetric partitioning, binary, or ternary, using a syntax such as flag.partitioning_symmetry or another flag.partitioning_method type of syntax. Also for the partitioning method and for either vertical or horizontal partitioning, another syntax to be encoded and signals may be a flag, e.g., flag.symmetry, indicating whether the non-QP technique will use symmetric or asymmetric partitioning (e.g., binary partitioning, ternary partitioning, the asymmetric partitioning shown in
For purposes of example,
For partitioning signaled as a partition in the vertical or horizontal direction with asymmetric partitioning, a syntax may further be signaled to indicate which child node is the larger child node and/or which child node (or child nodes in the case of ternary partitioning) is the smaller child node. The example in
In embodiments for which asymmetric partitioning is used, a flag may be signaled that identifies the ratio of the asymmetric split.
Similarly, the syntax for a symmetric ratio, such as flag.symmetric_ratio, may indicate a ratio for a ternary partitioned child node. The ratio may identify 3:1 if the larger child node is three parts relative the size of equal one-part child nodes. In embodiments, the ratio may identify 1:3:1 to specify that the middle node is the largest (3×). In embodiments, the ratio may identify a ratio, e.g., 3:1:1 and another syntax identifies which node is the largest. In embodiments, the ratio identifies more than two sizes, such as 3:2:1 or 4:3:1:1, depending on the split and whether a separate syntax is used to specify how the sizes are arranged within the coding unit.
For a 3:1 asymmetric partitioning such as that shown in
The restrictions may be imposed by the rules of encoding and decoding, such as if the larger child node resulting from a 3:1 asymmetric node is disallowed from further partitioning. For example, the smaller child node resulting from the Type 1 3:1 vertically partitioned node (shown as the furthest left branch in
After quadtree splitting, ternary tree splitting, and/or binary tree splitting using either QTBT or BTQT structures described above, the shapes represented by the QTBT'"'"'s final leaf nodes or BTQT'"'"'s final leaf nodes may represent the final CUs 102 to be coded, such as coding using inter prediction or intra prediction. In addition, the block structure may support the feature that luma and chroma have a separate tree structure. Currently, for P and B slices, the luma and chroma CTBs in one CTU share the same tree structure. For I slices, the luma CTB is partitioned into CUs by a tree structure, and the chroma CTBs are partitioned into chroma CUs by another QTBT structure. In one or more embodiments, a CU in an I slice may include a coding block of the luma component or coding blocks of two chroma components. In one or more embodiments, a CU in P and B slice a CU consists of coding blocks of all three colour components. In one or more embodiments, luma and chroma trees may be shared for an I slice.
In alternate embodiments WET can use a two-level coding block structure as an alternative to, or extension of, the QTBT partitioning described above. In the two-level coding block structure, a CTU 100 can first be partitioned at a high level into base units (BUs). The BUs can then be partitioned at a low level into operating units (OUs).
In embodiments employing the two-level coding block structure, at the high level a CTU 100 can be partitioned into BUs according to one of the QTBT structures described above, or according to a quadtree (QT) structure such as the one used in HEVC in which blocks can only be split into four equally sized sub-blocks. By way of a non-limiting example, a CTU 102 can be partitioned into BUs according to the QTBT structure described above with respect to
At the lower level in the two-level coding block structure, each BU partitioned from the CTU 100 can be further partitioned into one or more OUs. In some embodiments, when the BU is square, it can be split into OUs using quadtree partitioning or binary partitioning, such as symmetric or asymmetric binary partitioning. However, when the BU is not square, it can be split into OUs using binary partitioning only. Limiting the type of partitioning that can be used for non-square BUs can limit the number of bits used to signal the type of partitioning used to generate BUs.
Although the discussion below describes coding CUs 102, BUs and OUs can be coded instead of CUs 102 in embodiments that use the two-level coding block structure. By way of a non-limiting examples, BUs can be used for higher level coding operations such as intra prediction or inter prediction, while the smaller OUs can be used for lower level coding operations such as transforms and generating transform coefficients. Accordingly, syntax for be coded for BUs that indicate whether they are coded with intra prediction or inter prediction, or information identifying particular intra prediction modes or motion vectors used to code the BUs. Similarly, syntax for OUs can identify particular transform operations or quantized transform coefficients used to code the OUs.
Given a current CU 102, the encoder can obtain a prediction CU 702 either spatially using intra prediction at 704 or temporally using inter prediction at 706. The basic idea of prediction coding is to transmit a differential, or residual, signal between the original signal and a prediction for the original signal. At the receiver side, the original signal can be reconstructed by adding the residual and the prediction, as will be described below. Because the differential signal has a lower correlation than the original signal, fewer bits are needed for its transmission.
A sequence of coding units may make up a slice, and one or more slices may make up a picture. A slice may include one or more slice segments, each in its own NAL unit. A slice or slice segment may include header information for the slice or bitstream. A slice is a data structure that can be decoded independently from other slices of the same picture, in terms of entropy coding, signal prediction, and residual signal reconstruction. A slice can either be an entire picture or a region of a picture.
A slice, such as an entire picture or a portion of a picture, coded entirely with intra-predicted CUs 102 can be an I slice that can be decoded without reference to other slices, and as such can be a possible point where decoding can begin. A slice coded with at least some inter-predicted CUs can be a predictive (P) or bi-predictive (B) slice that can be decoded based on one or more reference pictures. P slices may use intra-prediction and inter-prediction with previously coded slices. For example, P slices may be compressed further than the I-slices by the use of inter-prediction, but need the coding of a previously coded slice to code them. B slices can use data from previous and/or subsequent slices for its coding, using intra-prediction or inter-prediction using an interpolated prediction from two different frames, thus increasing the accuracy of the motion estimation process. In some cases P slices and B slices can also or alternately be encoded using intra block copy, in which data from other portions of the same slice is used.
As will be discussed below, intra prediction or inter prediction can be performed based on reconstructed CUs 734 from previously coded CUs 102, such as neighboring CUs 102 or CUs 102 in reference pictures.
When a CU 102 is coded spatially with intra prediction at 704, an intra prediction mode can be found that best predicts pixel values of the CU 102 based on samples from neighboring CUs 102 in the picture.
When coding a CU'"'"'s luma component, the encoder can generate a list of candidate intra prediction modes. While HEVC had 35 possible intra prediction modes for luma components, in WET there are 67 possible intra prediction modes for luma components. These include a planar mode that uses a three dimensional plane of values generated from neighboring pixels, a DC mode that uses values averaged from neighboring pixels, and the 65 directional modes shown in
When generating a list of candidate intra prediction modes for a CU'"'"'s luma component, the number of candidate modes on the list can depend on the CU'"'"'s size. The candidate list can include: a subset of HEVC'"'"'s 35 modes with the lowest SATD (Sum of Absolute Transform Difference) costs; new directional modes added for WET that neighbor the candidates found from the HEVC modes; and modes from a set of six most probable modes (MPMs) for the CU 102 that are identified based on intra prediction modes used for previously coded neighboring blocks as well as a list of default modes.
When coding a CU'"'"'s chroma components, a list of candidate intra prediction modes can also be generated. The list of candidate modes can include modes generated with cross-component linear model projection from luma samples, intra prediction modes found for luma CBs in particular collocated positions in the chroma block, and chroma prediction modes previously found for neighboring blocks. The encoder can find the candidate modes on the lists with the lowest rate distortion costs, and use those intra prediction modes when coding the CU'"'"'s luma and chroma components. For example, an encoder rater-distortion optimization process of the QTBT structure may be used to determine the best block partitioning shape. Syntax can be coded in the bitstream that indicates the intra prediction modes used to code each CU 102.
After the best intra prediction modes for a CU 102 have been selected, the encoder can generate a prediction CU 402 using those modes. When the selected modes are directional modes, a 4-tap filter can be used to improve the directional accuracy. Columns or rows at the top or left side of the prediction block can be adjusted with boundary prediction filters, such as 2-tap or 3-tap filters.
The prediction CU 702 can be smoothed further with a position dependent intra prediction combination (PDPC) process that adjusts a prediction CU 702 generated based on filtered samples of neighboring blocks using unfiltered samples of neighboring blocks, or adaptive reference sample smoothing using 3-tap or 5-tap low pass filters to process reference samples.
When a CU 102 is coded temporally with inter prediction at 706, a set of motion vectors (MVs) can be found that points to samples in reference pictures that best predict pixel values of the CU 102. Inter prediction exploits temporal redundancy between slices by representing a displacement of a block of pixels in a slice. The displacement is determined according to the value of pixels in previous or following slices through a process called motion compensation. Motion vectors and associated reference indices that indicate pixel displacement relative to a particular reference picture can be provided in the bitstream to a decoder, along with the residual between the original pixels and the motion compensated pixels. The decoder can use the residual and signaled motion vectors and reference indices to reconstruct a block of pixels in a reconstructed slice.
In JVET, motion vector accuracy can be stored at 1/16 pel, and the difference between a motion vector and a CU'"'"'s predicted motion vector can be coded with either quarter-pel resolution or integer-pel resolution.
In JVET motion vectors can be found for multiple sub-CUs within a CU 102, using techniques such as advanced temporal motion vector prediction (ATMVP), spatial-temporal motion vector prediction (STMVP), affine motion compensation prediction, pattern matched motion vector derivation (PMMVD), and/or bi-directional optical flow (BIO).
Using ATMVP, the encoder can find a temporal vector for the CU 102 that points to a corresponding block in a reference picture. The temporal vector can be found based on motion vectors and reference pictures found for previously coded neighboring CUs 102. Using the reference block pointed to by a temporal vector for the entire CU 102, a motion vector can be found for each sub-CU within the CU 102.
STMVP can find motion vectors for sub-CUs by scaling and averaging motion vectors found for neighboring blocks previously coded with inter prediction, together with a temporal vector.
Affine motion compensation prediction can be used to predict a field of motion vectors for each sub-CU in a block, based on two control motion vectors found for the top corners of the block. For example, motion vectors for sub-CUs can be derived based on top corner motion vectors found for each 4×4 block within the CU 102.
PMMVD can find an initial motion vector for the current CU 102 using bilateral matching or template matching. Bilateral matching can look at the current CU 102 and reference blocks in two different reference pictures along a motion trajectory, while template matching can look at corresponding blocks in the current CU 102 and a reference picture identified by a template. The initial motion vector found for the CU 102 can then be refined individually for each sub-CU.
BIO can be used when inter prediction is performed with bi-prediction based on earlier and later reference pictures, and allows motion vectors to be found for sub-CUs based on the gradient of the difference between the two reference pictures.
In some situations local illumination compensation (LIC) can be used at the CU level to find values for a scaling factor parameter and an offset parameter, based on samples neighboring the current CU 102 and corresponding samples neighboring a reference block identified by a candidate motion vector. In JVET, the LIC parameters can change and be signaled at the CU level.
For some of the above methods the motion vectors found for each of a CU'"'"'s sub-CUs can be signaled to decoders at the CU level. For other methods, such as PMMVD and BIO, motion information is not signaled in the bitstream to save overhead, and decoders can derive the motion vectors through the same processes.
After the motion vectors for a CU 102 have been found, the encoder can generate a prediction CU 702 using those motion vectors. In some cases, when motion vectors have been found for individual sub-CUs, Overlapped Block Motion Compensation (OBMC) can be used when generating a prediction CU 702 by combining those motion vectors with motion vectors previously found for one or more neighboring sub-CUs.
When bi-prediction is used, JVET can use decoder-side motion vector refinement (DMVR) to find motion vectors. DMVR allows a motion vector to be found based on two motion vectors found for bi-prediction using a bilateral template matching process. In DMVR, a weighted combination of prediction CUs 702 generated with each of the two motion vectors can be found, and the two motion vectors can be refined by replacing them with new motion vectors that best point to the combined prediction CU 702. The two refined motion vectors can be used to generate the final prediction CU 702.
At 708, once a prediction CU 702 has been found with intra prediction at 704 or inter prediction at 706 as described above, the encoder can subtract the prediction CU 702 from the current CU 102 find a residual CU 710.
The encoder can use one or more transform operations at 712 to convert the residual CU 710 into transform coefficients 714 that express the residual CU 710 in a transform domain, such as using a discrete cosine block transform (DCT-transform) to convert data into the transform domain. WET allows more types of transform operations than HEVC, including DCT-II, DST-VII, DST-VII, DCT-VIII, DST-I, and DCT-V operations. The allowed transform operations can be grouped into sub-sets, and an indication of which sub-sets and which specific operations in those sub-sets were used can be signaled by the encoder. In some cases, large block-size transforms can be used to zero out high frequency transform coefficients in CUs 102 larger than a certain size, such that only lower-frequency transform coefficients are maintained for those CUs 102.
In some cases a mode dependent non-separable secondary transform (MDNSST) can be applied to low frequency transform coefficients 714 after a forward core transform. The MDNSST operation can use a Hypercube-Givens Transform (HyGT) based on rotation data. When used, an index value identifying a particular MDNSST operation can be signaled by the encoder.
At 716, the encoder can quantize the transform coefficients 714 into quantized transform coefficients 716. The quantization of each coefficient may be computed by dividing a value of the coefficient by a quantization step, which is derived from a quantization parameter (QP). In some embodiments, the Qstep is defined as 2(QP−4)/6. Because high precision transform coefficients 714 can be converted into quantized transform coefficients 716 with a finite number of possible values, quantization can assist with data compression. Thus, quantization of the transform coefficients may limit an amount of bits generated and sent by the transformation process. However, while quantization is a lossy operation, and the loss by quantization cannot be recovered, the quantization process presents a trade-off between quality of the reconstructed sequence and an amount of information needed to represent the sequence. For example, a lower QP value can result in better quality decoded video, although a higher amount of data may be required for representation and transmission. In contrast, a high QP value can result in lower quality reconstructed video sequences but with lower data and bandwidth needs.
NET can utilize variance-based adaptive quantization techniques, which allows every CU 102 to use a different quantization parameter for its coding process (instead of using the same frame QP in the coding of every CU 102 of the frame). The variance-based adaptive quantization techniques adaptively lowers the quantization parameter of certain blocks while increasing it in others. To select a specific QP for a CU 102, the CU'"'"'s variance is computed. In brief, if a CU'"'"'s variance is higher than the average variance of the frame, a higher QP than the frame'"'"'s QP may be set for the CU 102. If the CU 102 presents a lower variance than the average variance of the frame, a lower QP may be assigned.
At 720, the encoder can find final compression bits 722 by entropy coding the quantized transform coefficients 718. Entropy coding aims to remove statistical redundancies of the information to be transmitted. In NET, CABAC (Context Adaptive Binary Arithmetic Coding) can be used to code the quantized transform coefficients 718, which uses probability measures to remove the statistical redundancies. For CUs 102 with non-zero quantized transform coefficients 718, the quantized transform coefficients 718 can be converted into binary. Each bit (“bin”) of the binary representation can then be encoded using a context model. A CU 102 can be broken up into three regions, each with its own set of context models to use for pixels within that region.
Multiple scan passes can be performed to encode the bins. During passes to encode the first three bins (bin0, bin1, and bin2), an index value that indicates which context model to use for the bin can be found by finding the sum of that bin position in up to five previously coded neighboring quantized transform coefficients 718 identified by a template.
A context model can be based on probabilities of a bin'"'"'s value being ‘0’ or ‘1’. As values are coded, the probabilities in the context model can be updated based on the actual number of ‘0’ and ‘1’ values encountered. While HEVC used fixed tables to re-initialize context models for each new picture, in NET the probabilities of context models for new inter-predicted pictures can be initialized based on context models developed for previously coded inter-predicted pictures.
The encoder can produce a bitstream that contains entropy encoded bits 722 of residual CUs 710, prediction information such as selected intra prediction modes or motion vectors, indicators of how the CUs 102 were partitioned from a CTU 100 according to the QTBT structure, and/or other information about the encoded video. The bitstream can be decoded by a decoder as discussed below.
In addition to using the quantized transform coefficients 718 to find the final compression bits 722, the encoder can also use the quantized transform coefficients 718 to generate reconstructed CUs 734 by following the same decoding process that a decoder would use to generate reconstructed CUs 734. Thus, once the transformation coefficients have been computed and quantized by the encoder, the quantized transform coefficients 718 may be transmitted to the decoding loop in the encoder. After quantization of a CU'"'"'s transform coefficients, a decoding loop allows the encoder to generate a reconstructed CU 734 identical to the one the decoder generates in the decoding process. Accordingly, the encoder can use the same reconstructed CUs 734 that a decoder would use for neighboring CUs 102 or reference pictures when performing intra prediction or inter prediction for a new CU 102. Reconstructed CUs 102, reconstructed slices, or full reconstructed frames may serve as references for further prediction stages.
At the encoder'"'"'s decoding loop (and see below, for the same operations in the decoder) to obtain pixel values for the reconstructed image, a dequantization process may be performed. To dequantize a frame, for example, a quantized value for each pixel of a frame is multiplied by the quantization step, e.g., (Qstep) described above, to obtain reconstructed dequantized transform coefficients 726. For example, in the decoding process shown in
At 728, the dequantized transform coefficients 726 can be inverse transformed to find a reconstructed residual CU 730, such as by applying a DCT to the values to obtain the reconstructed image. At 732 the reconstructed residual CU 730 can be added to a corresponding prediction CU 702 found with intra prediction at 704 or inter prediction at 706, in order to find a reconstructed CU 734. While in some embodiments the encoder can perform intra prediction at 704 as described above, in other embodiments the encoder can follow the process disclosed herein for intra prediction template matching to generate a prediction CU 702 in the same way that a decoder would use template matching for intra prediction if information identifying the intra prediction mode used for the CU 102 is omitted from the bitstream.
At 736, one or more filters can be applied to the reconstructed data during the decoding process (in the encoder or, as described below, in the decoder), at either a picture level or CU level. For example, the encoder can apply a deblocking filter, a sample adaptive offset (SAO) filter, and/or an adaptive loop filter (ALF). The encoder'"'"'s decoding process may implement filters to estimate and transmit to a decoder the optimal filter parameters that can address potential artifacts in the reconstructed image. Such improvements increase the objective and subjective quality of the reconstructed video. In deblocking filtering, pixels near a sub-CU boundary may be modified, whereas in SAO, pixels in a CTU 100 may be modified using either an edge offset or band offset classification. JVET'"'"'s ALF can use filters with circularly symmetric shapes for each 2×2 block. An indication of the size and identity of the filter used for each 2×2 block can be signaled.
If reconstructed pictures are reference pictures, they can be stored in a reference buffer 738 for inter prediction of future CUs 102 at 706.
During the above steps, JVET allows a content adaptive clipping operations to be used to adjust color values to fit between lower and upper clipping bounds. The clipping bounds can change for each slice, and parameters identifying the bounds can be signaled in the bitstream.
As described above,
A WET decoder can receive a bitstream containing information about encoded CUs 102. The bitstream can indicate how CUs 102 of a picture were partitioned from a CTU 100 according to a QTBT structure. By way of a non-limiting example, the bitstream can identify how CUs 102 were partitioned from each CTU 100 in a QTBT using quadtree partitioning, symmetric binary partitioning, and/or asymmetric binary partitioning. The bitstream can also indicate prediction information for the CUs 102 such as intra prediction modes or motion vectors, and bits 902 representing entropy encoded residual CUs. In some embodiments the encoder can have omitted information in the bitstream about intra prediction modes used to encode some or all CUs 102 coded using intra prediction, and as such the decoder can use template matching for intra prediction as described below with respect to the following figures.
At 904 the decoder can decode the entropy encoded bits 902 using the CABAC context models signaled in the bitstream by the encoder. The decoder can use parameters signaled by the encoder to update the context models'"'"' probabilities in the same way they were updated during encoding.
After reversing the entropy encoding at 904 to find quantized transform coefficients 906, the decoder can dequantize them at 908 to find dequantized transform coefficients 910. If an MDNSST operation was performed during encoding, that operation can be reversed by the decoder after dequantization.
At 912, the dequantized transform coefficients 910 can be inverse transformed to find a reconstructed residual CU 914. At 916, the reconstructed residual CU 914 can be added to a corresponding prediction CU 926 found with intra prediction at 922 or inter prediction at 924, in order to find a reconstructed CU 918.
At 920, one or more filters can be applied to the reconstructed data, at either a picture level or CU level. For example, the decoder can apply a deblocking filter, a sample adaptive offset (SAO) filter, and/or an adaptive loop filter (ALF). As described above, the in-loop filters located in the decoding loop of the encoder may be used to estimate optimal filter parameters to increase the objective and subjective quality of a frame. These parameters are transmitted to the decoder to filter the reconstructed frame at 920 to match the filtered reconstructed frame in the encoder.
After reconstructed pictures have been generated by finding reconstructed CUs 918 and applying signaled filters, the decoder can output the reconstructed pictures as output video 928. If reconstructed pictures are to be used as reference pictures, they can be stored in a reference buffer 930 for inter prediction of future CUs 102 at 924.
In step 1008, the quantized transform coefficients 906 associated with the encoded bitstream 902 can be determined and de-quantized transform coefficients 910 can then be determined from the quantized transform coefficients 906 in step 1010.
In step 1012, it can be determined whether an MDNSST operation was performed during encoding and/or if the bitstream 902 contains indications that an MDNSST operation was applied to the bitstream 902. If it is determined that an MDNSST operation was performed during the encoding process or the bitstream 902 contains indications that an MDNSST operation was applied to the bitstream 902, then an inverse MDNSST operation 1014 can be implemented before an inverse transform operation 912 is performed on the bitstream 902 in step 1016. Alternately, an inverse transform operation 912 can be performed on the bitstream 902 in step 1016 absent application of an inverse MDNSST operation in step 1014. The inverse transform operation 912 in step 1016 can determine and/or construct a reconstructed residual CU 914.
In step 1018, the reconstructed residual CU 914 from step 1016 can be combined with a prediction CU 918. The prediction CU 918 can be one of an intra-prediction CU 922 determined in step 1020 and an inter-prediction unit 924 determined in step 1022.
In step 1024, any one or more filters 920 can be applied to the reconstructed CU 914 and output in step 1026. In some embodiments filters 920 may not be applied in step 1024.
In some embodiments, in step 1028, the reconstructed CU 918 can be stored in a reference buffer 930.
The execution of the sequences of instructions required to practice the embodiments can be performed by a computer system 1100 as shown in
A computer system 1100 according to an embodiment will now be described with reference to
Each computer system 1100 can include a communication interface 1114 coupled to the bus 1106. The communication interface 1114 provides two-way communication between computer systems 1100. The communication interface 1114 of a respective computer system 1100 transmits and receives electrical, electromagnetic or optical signals, which include data streams representing various types of signal information, e.g., instructions, messages and data. A communication link 1115 links one computer system 1100 with another computer system 1100. For example, the communication link 1115 can be a LAN, in which case the communication interface 1114 can be a LAN card, or the communication link 1115 can be a PSTN, in which case the communication interface 1114 can be an integrated services digital network (ISDN) card or a modem, or the communication link 1115 can be the Internet, in which case the communication interface 1114 can be a dial-up, cable or wireless modem.
A computer system 1100 can transmit and receive messages, data, and instructions, including program, i.e., application, code, through its respective communication link 1115 and communication interface 1114. Received program code can be executed by the respective processor(s) 1107 as it is received, and/or stored in the storage device 1110, or other associated non-volatile media, for later execution.
In an embodiment, the computer system 1100 operates in conjunction with a data storage system 1131, e.g., a data storage system 1131 that contains a database 1132 that is readily accessible by the computer system 1100. The computer system 1100 communicates with the data storage system 1131 through a data interface 1133. A data interface 1133, which is coupled to the bus 1106, transmits and receives electrical, electromagnetic or optical signals, which include data streams representing various types of signal information, e.g., instructions, messages and data. In embodiments, the functions of the data interface 1133 can be performed by the communication interface 1114.
Computer system 1100 includes a bus 1106 or other communication mechanism for communicating instructions, messages and data, collectively, information, and one or more processors 1107 coupled with the bus 1106 for processing information. Computer system 1100 also includes a main memory 1108, such as a random access memory (RAM) or other dynamic storage device, coupled to the bus 1106 for storing dynamic data and instructions to be executed by the processor(s) 1107. The main memory 1108 also can be used for storing temporary data, i.e., variables, or other intermediate information during execution of instructions by the processor(s) 1107.
The computer system 1100 can further include a read only memory (ROM) 1109 or other static storage device coupled to the bus 1106 for storing static data and instructions for the processor(s) 1107. A storage device 1110, such as a magnetic disk or optical disk, can also be provided and coupled to the bus 1106 for storing data and instructions for the processor(s) 1107.
A computer system 1100 can be coupled via the bus 1106 to a display device 1111, such as, but not limited to, a cathode ray tube (CRT) or a liquid-crystal display (LCD) monitor, for displaying information to a user. An input device 1112, e.g., alphanumeric and other keys, is coupled to the bus 1106 for communicating information and command selections to the processor(s) 1107.
According to one embodiment, an individual computer system 1100 performs specific operations by their respective processor(s) 1107 executing one or more sequences of one or more instructions contained in the main memory 1108. Such instructions can be read into the main memory 1108 from another computer-usable medium, such as the ROM 1109 or the storage device 1110. Execution of the sequences of instructions contained in the main memory 1108 causes the processor(s) 1107 to perform the processes described herein. In alternative embodiments, hard-wired circuitry can be used in place of or in combination with software instructions. Thus, embodiments are not limited to any specific combination of hardware circuitry and/or software.
The term “computer-usable medium,” as used herein, refers to any medium that provides information or is usable by the processor(s) 1107. Such a medium can take many forms, including, but not limited to, non-volatile, volatile and transmission media. Non-volatile media, i.e., media that can retain information in the absence of power, includes the ROM 1109, CD ROM, magnetic tape, and magnetic discs. Volatile media, i.e., media that cannot retain information in the absence of power, includes the main memory 1108. Transmission media includes coaxial cables, copper wire and fiber optics, including the wires that comprise the bus 1106. Transmission media can also take the form of carrier waves; i.e., electromagnetic waves that can be modulated, as in frequency, amplitude or phase, to transmit information signals. Additionally, transmission media can take the form of acoustic or light waves, such as those generated during radio wave and infrared data communications. In the foregoing specification, the embodiments have been described with reference to specific elements thereof. It will, however, be evident that various modifications and changes can be made thereto without departing from the broader spirit and scope of the embodiments. For example, the reader is to understand that the specific ordering and combination of process actions shown in the process flow diagrams described herein is merely illustrative, and that using different or additional process actions, or a different combination or ordering of process actions can be used to enact the embodiments. The specification and drawings are, accordingly, to be regarded in an illustrative rather than restrictive sense
It should also be noted that the present invention can be implemented in a variety of computer systems. The various techniques described herein can be implemented in hardware or software, or a combination of both. Preferably, the techniques are implemented in computer programs executing on programmable computers that each include a processor, a storage medium readable by the processor (including volatile and non-volatile memory and/or storage elements), at least one input device, and at least one output device. Program code is applied to data entered using the input device to perform the functions described above and to generate output information. The output information is applied to one or more output devices. Each program is preferably implemented in a high level procedural or object oriented programming language to communicate with a computer system. However, the programs can be implemented in assembly or machine language, if desired. In any case, the language can be a compiled or interpreted language. Each such computer program is preferably stored on a storage medium or device (e.g., ROM or magnetic disk) that is readable by a general or special purpose programmable computer for configuring and operating the computer when the storage medium or device is read by the computer to perform the procedures described above. The system can also be considered to be implemented as a computer-readable storage medium, configured with a computer program, where the storage medium so configured causes a computer to operate in a specific and predefined manner. Further, the storage elements of the exemplary computing applications can be relational or sequential (flat file) type computing databases that are capable of storing data in various combinations and configurations.
Destination device 14 may receive encoded video data from source device 12 via a channel 16. Channel 16 may comprise a type of medium or device capable of moving the encoded video data from source device 12 to destination device 14. In one example, channel 16 may comprise a communication medium that enables source device 12 to transmit encoded video data directly to destination device 14 in real-time.
In this example, source device 12 may modulate the encoded video data according to a communication standard, such as a wireless communication protocol, and may transmit the modulated video data to destination device 14. The communication medium may comprise a wireless or wired communication medium, such as a radio frequency (RF) spectrum or one or more physical transmission lines. The communication medium may form part of a packet-based network, such as a local area network, a wide-area network, or a global network such as the Internet. The communication medium may include routers, switches, base stations, or other equipment that facilitates communication from source device 12 to destination device 14. In another example, channel 16 may correspond to a storage medium that stores the encoded video data generated by source device 12.
In the example of
Video encoder 20 may encode the captured, pre-captured, or computer-generated video data. An input image may be received by the video encoder 20 and stored in the input frame memory 21. The general purpose processor 23 may load information from here and perform encoding. The program for driving the general purpose processor may be loaded from a storage device, such as the example memory modules depicted in
The video encoder 20 may include a resampling module 25 which may be configured to code (e.g., encode) video data in a scalable video coding scheme that defines at least one base layer and at least one enhancement layer. Resampling module 25 may resample at least some video data as part of an encoding process, wherein resampling may be performed in an adaptive manner using resampling filters.
The encoded video data, e.g., a coded bit stream, may be transmitted directly to destination device 14 via output interface 28 of source device 12. In the example of
The encoded video data may also be stored onto a storage medium or a file server for later access by destination device 14 for decoding and/or playback. For example, the coded bitstream may be temporarily stored in the input buffer 31, then loaded in to the general purpose processor 33. The program for driving the general purpose processor may be loaded from a storage device or memory. The general purpose processor may use a process memory 32 to perform the decoding. The video decoder 30 may also include a resampling module 35 similar to the resampling module 25 employed in the video encoder 20.
Display device 38 may be integrated with or may be external to destination device 14. In some examples, destination device 14 may include an integrated display device and may also be configured to interface with an external display device. In other examples, destination device 14 may be a display device. In general, display device 38 displays the decoded video data to a user.
Video encoder 20 and video decoder 30 may operate according to a video compression standard. ITU-T VCEG (Q6/16) and ISO/IEC MPEG (JTC 1/SC 29/WG 11) are studying the potential need for standardization of future video coding technology with a compression capability that significantly exceeds that of the current High Efficiency Video Coding HEVC standard (including its current extensions and near-term extensions for screen content coding and high-dynamic-range coding). The groups are working together on this exploration activity in a joint collaboration effort known as the Joint Video Exploration Team (WET) to evaluate compression technology designs proposed by their experts in this area. A recent capture of JVET development is described in the “Algorithm Description of Joint Exploration Test Model 5 (JEM 5)”, WET-E1001-V2, authored by J. Chen, E. Alshina, G. Sullivan, J. Ohm, J. Boyce.
Additionally or alternatively, video encoder 20 and video decoder 30 may operate according to other proprietary or industry standards that function with the disclosed JVET features. Thus, other standards such as the ITU-T H.264 standard, alternatively referred to as MPEG-4, Part 10, Advanced Video Coding (AVC), or extensions of such standards. Thus, while newly developed for JVET, techniques of this disclosure are not limited to any particular coding standard or technique. Other examples of video compression standards and techniques include MPEG-2, ITU-T H.263 and proprietary or open source compression formats and related formats.
Video encoder 20 and video decoder 30 may be implemented in hardware, software, firmware or any combination thereof. For example, the video encoder 20 and decoder 30 may employ one or more processors, digital signal processors (DSPs), application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), discrete logic, or any combinations thereof. When the video encoder 20 and decoder 30 are implemented partially in software, a device may store instructions for the software in a suitable, non-transitory computer-readable storage medium and may execute the instructions in hardware using one or more processors to perform the techniques of this disclosure. Each of video encoder 20 and video decoder 30 may be included in one or more encoders or decoders, either of which may be integrated as part of a combined encoder/decoder (CODEC) in a respective device.
Aspects of the subject matter described herein may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer, such as the general purpose processors 23 and 33 described above. Generally, program modules include routines, programs, objects, components, data structures, and so forth, which perform particular tasks or implement particular abstract data types. Aspects of the subject matter described herein may also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules may be located in both local and remote computer storage media including memory storage devices.
Examples of memory include random access memory (RAM), read only memory (ROM), or both. Memory may store instructions, such as source code or binary code, for performing the techniques described above. Memory may also be used for storing variables or other intermediate information during execution of instructions to be executed by a processor, such as processor 23 and 33.
A storage device may also store instructions, instructions, such as source code or binary code, for performing the techniques described above. A storage device may additionally store data used and manipulated by the computer processor. For example, a storage device in a video encoder 20 or a video decoder 30 may be a database that is accessed by computer system 23 or 33. Other examples of storage device include random access memory (RAM), read only memory (ROM), a hard drive, a magnetic disk, an optical disk, a CD-ROM, a DVD, a flash memory, a USB memory card, or any other medium from which a computer can read.
A memory or storage device may be an example of a non-transitory computer-readable storage medium for use by or in connection with the video encoder and/or decoder. The non-transitory computer-readable storage medium contains instructions for controlling a computer system to be configured to perform functions described by particular embodiments. The instructions, when executed by one or more computer processors, may be configured to perform that which is described in particular embodiments.
Also, it is noted that some embodiments have been described as a process which can be depicted as a flow diagram or block diagram. Although each may describe the operations as a sequential process, many of the operations can be performed in parallel or concurrently. In addition, the order of the operations may be rearranged. A process may have additional steps not included in the figures.
Particular embodiments may be implemented in a non-transitory computer-readable storage medium for use by or in connection with the instruction execution system, apparatus, system, or machine. The computer-readable storage medium contains instructions for controlling a computer system to perform a method described by particular embodiments. The computer system may include one or more computing devices. The instructions, when executed by one or more computer processors, may be configured to perform that which is described in particular embodiments
As used in the description herein and throughout the claims that follow, “a”, “an”, and “the” includes plural references unless the context clearly dictates otherwise. Also, as used in the description herein and throughout the claims that follow, the meaning of “in” includes “in” and “on” unless the context clearly dictates otherwise.
Although the subject matter has been described in language specific to structural features and/or methodological acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above.
Although embodiments have been disclosed herein in detail and in language specific to structural features and/or methodological acts above, it is to be understood that those skilled in the art will readily appreciate that many additional modifications are possible in the exemplary embodiments without materially departing from the novel teachings and advantages of the invention. Moreover, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features or acts described above. Accordingly, these and all such modifications are intended to be included within the scope of this invention construed in breadth and scope in accordance with the appended claims.