Digital video compression system utilizing vector adaptive transform

• US 5,367,629 A
• Filed: 12/18/1992
• Issued: 11/22/1994
• Est. Priority Date: 12/18/1992
• Status: Expired due to Term

• Current Assignee: ShareVision Technology, Inc. (Creative Technology Ltd.)
• Original Assignee: ShareVision Technology, Inc. (Creative Technology Ltd.)
• Inventors: Chu, Frank J., Yeh, C. Lung
• Primary Examiner(s): RICHARDSON, ROBERT
• Application Number:
  US07/993,039
• Time in Patent Office: 704 Days
• Field of Search: 358/426, 358/133, 358/433, 358/470, 358/432, 358/427, 358/160, 358/166, 345/189, 345/202, 395/162-166, 382/56, 348/384, 348/403, 348/405, 348/406, 348/409, 348/415
• US Class Current: 382/253
• CPC Class Codes: G06T 9/008 Vector quantisation
• CPC Class Codes: H04N 19/117 Filters, e.g. for pre-processing or post-processing sub-band filter banks H04N19/635
• CPC Class Codes: H04N 19/124 Quantisation
• CPC Class Codes: H04N 19/126 Details of normalisation or weighting functions, e.g. normalisation matrices or variable uniform quantisers
• CPC Class Codes: H04N 19/129 Scanning of coding units, e.g. zig-zag scan of transform coefficients or flexible macroblock ordering [FMO]
• CPC Class Codes: H04N 19/146 Data rate or code amount at the encoder output
• CPC Class Codes: H04N 19/152 by measuring the fullness of the transmission buffer
• CPC Class Codes: H04N 19/176 the region being a block, e.g. a macroblock
• CPC Class Codes: H04N 19/18 the unit being a set of transform coefficients
• CPC Class Codes: H04N 19/192 the adaptation method, adaptation tool or adaptation type being iterative or recursive
• CPC Class Codes: H04N 19/523 with sub-pixel accuracy
• CPC Class Codes: H04N 19/61 in combination with predictive coding

Abstract

A video compression system comprises a pre-processing section, an encoder, and a post-processing section. The pre-processing section employs a median decimation filter which combines median filtering and decimation process. The preprocessing section also employs adaptive temporal filtering and content adaptive noise reduction filtering to provide images with proper smoothness and sharpness to match the encoder characteristics. The encoder employs a two pass look-ahead allocation rate buffer control scheme where the numbers of bits allocated and subsequently generated for each block may differ. In the first pass, the means square error for each block is estimated to determine the number of bits assigned to each block in a frame, In the second pass, the degree of compression is controlled as a function of the total number of bits generated for all the preceding blocks and the sum of the bits allocated to such preceding blocks. The DCT coefficients are vector coded by scanning the coefficients using three different scan paths, and the one providing the most compact path is chosen. Each scan vector is divided into eight zones and coded using variable length coding. An all vector module is employed to take advantage of the low rate coding. Background information is employed in addition to the previous frame for prediction to achieve better coding gain. The post-processing section filters block boundaries to remove blocking artifacts. Depending on the image characteristic (average quantization applied), different kinds of filtering is applied to enhance the images.

Claims

1. A method for encoding frames of digitized video information, wherein the frames are divided into blocks, and wherein numbers of bits available for encoding video information change over time so that the number of bits available for encoding a frame may differ from the number of bits available for encoding a subsequent frame, said method comprising:

performing a spatial domain to transform domain transformation of video information related to each block of a frame to obtain a corresponding block of transform coefficients;

ascertaining the number of bits available for encoding said frame;

allocating a number of bits to the encoding of each block of transform coefficients for all the blocks in said frame; and

quantizing and coding each block of transform coefficients to generate variable length coding bits for each block, wherein the number of bits generated for each block is a function of the number of bits allocated to the encoding of a different block.

2. The method of claim 1, wherein said blocks of the frame are encoded sequentially in a time sequence, wherein the number of bits generated in the quantizing and coding of a block of transform coefficients is a function of the number of bits allocated to the encoding of a subsequent block in the time sequence in a look-ahead bit allocation scheme.

3. The method of claim 2, said method further comprising:

prior to quantizing and coding a block of transform coefficients, determining the sum total number of bits generated in quantizing and coding all blocks of transform coefficients that precede such block in the time sequence, if any;

wherein said step of quantizing such block of transform coefficients uses information regarding the sum total number of bits allocated to all the blocks of the frame that have been quantized and coded up to such quantizing step, and the sum total number of bits generated in quantizing and coding all the preceding blocks of transform coefficients, to control at such quantizing step the number of variable length coding bits generated for such block.

4. An apparatus for encoding frames of digitized video information, wherein the frames are divided into blocks, and wherein numbers of bits available for encoding video information change over time so that the number of bits available for encoding a frame may differ from the number of bits available for encoding a subsequent frame, said apparatus comprising:

means for performing a spatial domain to transform domain transformation of video information related to each block of a frame to obtain a corresponding block of transform coefficients;

means for ascertaining the number of bits available for encoding said frame;

means for allocating a number of bits to the encoding of each block of transform coefficients for all the blocks in said frame; and

means for quantizing and coding each block of transform coefficients to generate variable length coding bits for each block, wherein the number of bits generated is a function of the number of bits allocated to the encoding of a different block of transform coefficients.

5. A method for encoding frames of pixel values of video information, wherein the frames are divided into blocks, said method comprising:

performing a spatial domain to transform domain transformation of video information related to each block of a frame to obtain a corresponding block of transform coefficients;

quantizing and coding each block of transform coefficients to generate variable length coding bits for each block, wherein the quantizing step is performed by reference to one index within a set of numerical indices indicating number of bits resulting from the quantizing step, and wherein said indices have values that vary directly with the number of bits resulting from the quantizing step;

computing an average value of the indices by reference to which the blocks of a frame have been quantized and coded in the quantizing and coding step;

comparing said average value to predetermined range values; and

multiplying each pixel value by a selected matrix element when the average value is within predetermined range values.

6. A method for encoding frames of digitized video information, wherein the frames are divided into blocks, said method comprising:

performing a spatial domain to transform domain transformation of video information related to each block of a frame to obtain a corresponding block of transform coefficients;

quantizing and coding each block of transform coefficients to generate variable length coding bits for each block, wherein said coding step includes;

first scanning at least some of the coefficients of each block along at least two different paths and counting the number of nonzero scanned coefficients along each path to obtain at least two numbers;

comparing the numbers obtained to find the largest of the numbers; and

scanning the coefficients of each block along the path that results in the largest number before coding each block of transform coefficients.

7. The method of claim 6, further comprising:

dividing each block into two or more sets of zones each corresponding to a scan path, wherein said first scanning step scans only the coefficients in one of the zones along said different paths.

8. A method for encoding frames of digitized video information, wherein the frames are divided into blocks, said method comprising:

performing a spatial domain to transform domain transformation of video information related to each block of a frame to obtain a corresponding block of transform coefficients;

quantizing and coding each block of transform coefficients to generate variable length coding bits for each block;

dividing each block into two or more sets of zones; and

forming a scan vector of at least one zone and variable length coding of coefficients of said at least one zone.

9. The method of claim 8, further comprising;

detecting whether the remaining coefficients are zero after said forming step; and

sending end of block when the remaining coefficients are detected to be zero after said forming step.

10. The method of claim 8, wherein said forming step also detects whether the vector formed contains all 1'"'"'s or 0'"'"'s.

11. A method for encoding frames of video information, said frames divided into blocks, wherein the frames are encoded sequentially in a time sequence, so that each block in a frame has a corresponding block in a different frame, said method comprising:

performing a spatial domain to transform domain transformation of video information related to each block of a frame to obtain a corresponding block of transform coefficients;

quantizing and coding each block of transform coefficients to generate variable length coding bits for each block;

reconstructing each block from its variable length coding bits;

storing said reconstructed block;

storing a predetermined frame of video information as the background frame, said frame containing a background block corresponding to said reconstructed block;

comparing a current block to be quantized and coded to said reconstructed block and to the background block to determine a prediction block; and

using the prediction block to derive said video information related to each block that is transformed in said performing step.