Flexible quantization

• US 8,711,925 B2
• Filed: 05/05/2006
• Issued: 04/29/2014
• Est. Priority Date: 05/05/2006
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Tu, Chengjie, Srinivasan, Sridhar
• Primary Examiner(s): Perungavoor, Sath V
• Assistant Examiner(s): PONTIUS, JAMES M
• Application Number:
  US11/418,690
• Publication Number:
  US 20070258518A1
• Time in Patent Office: 2,916 Days
• Field of Search: 375/240.03
• US Class Current: 375/240.03
• 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 19/136 Incoming video signal characteristics or properties
• CPC Class Codes: H04N 19/176 the region being a block, e.g. a macroblock
• CPC Class Codes: H04N 19/46 Embedding additional information in the video signal during the compression process H04N19/517, H04N19/68, H04N19/70 take precedence
• CPC Class Codes: H04N 19/61 in combination with predictive coding

Abstract

A digital media encoder/decoder uses a flexible quantization technique that provides the ability to vary quantization along various dimensions of the encoded digital media data, including spatial, frequency sub bands and color channels. The codec utilizes a signaling scheme to signal various permutations of flexible quantization combinations efficiently for primary usage scenarios. When a choice of quantizer is available, the codec efficiently encodes the current quantizer by defining a subset of quantizers and indexes the current quantizer from the set.

Claims

1. A method of decoding digital media, comprising:

with a computer,

receiving compressed digital image data, the compressed digital image data having been partitioned along more than one dimension;

determining multiple sets of quantization parameter indices from signals in the compressed digital image data, the compressed digital image data having been quantized across more than one partitioned dimension of the digital image data, the signals indicating whether quantization parameter indices used during decoding vary within a frequency band dimension and within a color channel dimension, the quantization parameter indices available across the frequency band dimension including, for a given set of the multiple sets of quantization parameter indices, a first quantization parameter index for one or more DC components, a second quantization parameter index for multiple low-pass components, and a third quantization parameter index for multiple high-pass components;

for each of multiple macroblocks of the compressed digital image data;

selecting one of the multiple sets of quantization parameter indices based on a value signaled for the macroblock; and

determining one or more quantization parameters for the selected set of quantization parameter indices;

dequantizing the compressed digital image data in accordance with the determined quantization parameters;

applying an inverse of a block transform to blocks of the dequantized digital image data; and

outputting decompressed digital image data,

wherein;

(a) the multiple high-pass coefficients are AC coefficients from first transforms of values of blocks of a macroblock;

(b) the multiple low-pass coefficients are AC coefficients from a second transform of DC coefficients from the first transforms; and

(c) the one or more DC coefficients include a DC coefficient from the second transform of the DC coefficients from the first transforms.

2. The method of claim 1, wherein the signals further include (i) a signal indicating whether the low-pass components are to be decoded using the first quantization parameter index for the DC components and (ii) a signal indicating whether the high-pass components are to be quantized using the second quantization parameter index for low-pass components.

3. The method of claim 1, wherein the signals further indicate whether quantization parameters used during dequantization vary within a color channel dimension, and wherein the signals further include a signal indicating (i) whether to dequantize DC components of a luminance channel and both chrominance channels uniformly, (ii) whether to dequantize the DC component of the luminance channel separately from the DC components of the chrominance channels, or (iii) whether to dequantize the DC component of the luminance channel and each of the DC components for the chrominance channels independently.

4. The method of claim 1, wherein said determining quantization parameters further comprises:

determining whether to specify DC quantization parameters (a) uniformly for a luminance channel and all chrominance channels;

(b) separately between (i) the luminance channel and (ii) all chrominance channels;

or (c) independently for the luminance channel and each of the chrominance channels.

5. The method of claim 4, wherein said determining quantization parameters further comprises:

determining whether to specify LP quantization parameters using the DC quantization parameters and, if not, determining whether to specify the LP quantization parameters (a) uniformly for a luminance channel and all chrominance channels;

(b) separately between (i) the luminance channel and (ii) all chrominance channels;

or (c) independently for the luminance channel and each of the chrominance channels.

6. The method of claim 1 wherein the signals include a flag that indicates whether the second quantization parameter index is same as the third quantization parameter index.

7. A digital media processing device comprising:

a memory for storing digital media data; and

a digital media processor programmed to encode and/or decode the digital media data by applying forward or inverse operations of a transform to blocks of the digital media data, and applying quantizers to the blocks of digital media data, wherein the quantizers include a set of frequency band quantizers, the set of frequency band quantizers including;

(a) an HP quantizer for multiple high-pass coefficients of a macroblock, wherein the multiple high-pass coefficients are AC coefficients from first transforms of values of blocks of the macroblock;

(b) an LP quantizer for multiple low-pass coefficients of the macroblock, wherein the multiple low-pass coefficients are AC coefficients from a second transform of DC coefficients from the first transforms; and

(c) a DC quantizer for a DC coefficient of the macroblock, wherein the DC coefficient is a DC coefficient from the second transform of the DC coefficients from the first transforms.

8. The device of claim 7 wherein the set of frequency band quantizers are controlled by signals in the bitstream that allow variation of quantization spatially and across color channels.

9. At least one memory device or magnetic disk having program code stored thereon for causing a digital media processing device to perform a method of processing digital media data according to a codec, the method comprising:

selecting combinations of available quantization modes to be applied to the digital media data, the available quantization modes including at least a first quantization mode in which low pass coefficients use the same quantizer as a DC coefficient and a second quantization mode in which high pass coefficients use the same quantizer as low pass coefficients, the first quantization mode being signaled by a first syntax element, the second quantization mode being signaled by a second syntax element;

applying a two-stage transform to blocks of the digital media data;

quantizing the digital media data according to the selected combinations of quantization modes;

encoding the quantized digital media data in a compressed data stream; and

signaling the selected combinations of quantization modes in the compressed data stream, the compressed data stream including the first syntax element and the second syntax element,

wherein the high pass coefficients are AC coefficients from first transforms of values of the blocks, and wherein the low pass coefficients are AC coefficients from a second transform of DC coefficients from the first transforms.

10. The at least one memory device or magnetic disk of claim 9 wherein the available quantization modes further comprise a mode for uniform quantization per frame of the digital media data, a mode for uniform quantization per tile of the digital media data, and a mode for independently quantizing each macroblock of the digital media data.

11. The at least one memory device or magnetic disk of claim 9 wherein the available quantization modes further comprise a mode for uniform quantization across color channels of the digital media data, a mode for partially uniformly quantizing across the color channels of the digital media data, and a mode for independently quantizing the color channels of the digital media data.

12. The at least one memory device or magnetic disk of claim 9 wherein the method further comprises defining a set of available quantizers and specifying a current quantizer by indexing from the set.

13. At least one memory device or magnetic disk having program code stored thereon for causing a digital media processing device to perform a method of processing digital media data according to a codec, the method comprising decoding a compressed data stream encoded according to the method of claim 9.

14. At least one memory device or magnetic disk having program code stored thereon for causing a digital media processing device to perform a method of processing digital media data according to a codec, the method comprising:

receiving compressed digital image data;

determining multiple sets of quantization parameter indices from signals in the compressed digital image data, the signals indicating whether quantization parameter indices used during decoding vary within a frequency band dimension and within a color channel dimension, the quantization parameter indices available across the frequency band dimension including, for a given set of the multiple sets of quantization parameter indices, a first quantization parameter index for one or more DC components, a second quantization parameter index for multiple low-pass components, and a third quantization parameter index for multiple high-pass components;

for each of multiple macroblocks of the compressed digital image data;

selecting one of the multiple sets of quantization parameter indices based on a value signaled for the macroblock; and

determining one or more quantization parameters for the selected set of quantization parameter indices;

dequantizing the compressed digital image data in accordance with the determined quantization parameters;

applying an inverse of a block transform to blocks of the dequantized digital image data; and

outputting decompressed digital image data,

wherein (a) the multiple high-pass coefficients are AC coefficients from first transforms of values of blocks of a macroblock;

(b) the multiple low-pass coefficients are AC coefficients from a second transform of DC coefficients from the first transforms; and

(c) the one or more DC coefficients include a DC coefficient from the second transform of the DC coefficients from the first transforms.

15. The at least one memory device or magnetic disk of claim 14, wherein the signals further indicate whether quantization parameters used during dequantization vary within a color channel dimension, and wherein said determining quantization parameters comprises:

determining whether to specify DC quantization parameters (a) uniformly for a luminance channel and all chrominance channels;

(b) separately between (i) the luminance channel and (ii) all chrominance channels;

or (c) independently for the luminance channel and each of the chrominance channels.

16. The at least one memory device or magnetic disk of claim 15, wherein said determining quantization parameters further comprises:

determining whether to specify LP quantization parameters using the DC quantization parameters and, if not, determining whether to specify the LP quantization parameters (a) uniformly for a luminance channel and all chrominance channels;

(b) separately between (i) the luminance channel and (ii) all chrominance channels;

or (c) independently for the luminance channel and each of the chrominance channels.

17. The at least one memory device or magnetic disk of claim 14, wherein the signals include a flag that indicates whether the second quantization parameter index is same as the third quantization parameter index.