COMPUTATIONAL COMPLEXITY AND PRECISION CONTROL IN TRANSFORM-BASED DIGITAL MEDIA CODEC

• US 20080198935A1
• Filed: 06/29/2007
• Published: 08/21/2008
• Est. Priority Date: 02/21/2007
• Status: Active Grant

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Srinivasan, Sridhar, Regunathan, Shankar, Tu, Chengjie
• Granted Patent: US 8,942,289 B2
• Time in Patent Office: Days
• US Class Current: 375/240.23
• CPC Class Codes: H04N 19/122 Selection of transform size, e.g. 8x8 or 2x4x8 DCT; Selection of sub-band transforms of varying structure or type
• CPC Class Codes: H04N 19/156 Availability of hardware or computational resources, e.g. encoding based on power-saving criteria
• 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 19/60 using transform coding
• CPC Class Codes: H04N 21/6547 comprising parameters, e.g. for client setup

Abstract

A digital media encoder/decoder includes signaling of various modes relating to computation complexity and precision at decoding. The encoder may send a syntax element indicating arithmetic precision (e.g., using 16 or 32-bit operations) of the transform operations performed at decoding. The encoder also may signal whether to apply scaling at the decoder output, which permits a wider dynamic range of intermediate data at decoding, but adds to computational complexity due to the scaling operation.

Claims

1. A digital media decoding method comprising:

receiving a compressed digital media bitstream at a digital media decoder;

parsing a syntax element from the bitstream signaling a degree of arithmetic precision to use for transform computations during processing of the digital media data; and

outputting a reconstructed image.

2. The digital media decoding method of claim 1 wherein the syntax element signals to use one of a high arithmetic precision or a low arithmetic precision.

3. The digital media decoding method of claim 2 wherein the high arithmetic precision is 32-bit number processing, and the low arithmetic precision is 16-bit number processing.

4. The digital media decoding method of claim 2 further comprising:

decoding blocks of transform coefficients from the compressed digital media bitstream;

in the case that the syntax element signals use of the high arithmetic precision, applying an inverse transform to the transform coefficients using high arithmetic precision processing; and

in the case that the syntax element signals use of the low arithmetic precision, applying an inverse transform to the transform coefficients using low arithmetic precision processing.

5. The digital media decoding method of claim 4 wherein the high arithmetic precision is 32-bit number processing, and the low arithmetic precision is 16-bit number processing.

6. The digital media decoding method of claim 2 further comprising:

decoding blocks of transform coefficients from the compressed digital media bitstream;

applying an inverse transform to the transform coefficients using high arithmetic precision processing regardless of the degree of arithmetic precision signaled via the syntax element.

7. A digital media encoding method comprising:

receiving digital media data at a digital media encoder;

making a decision whether to use lower precision arithmetic for transform computations during processing of the digital media data;

representing the decision whether to use the lower precision arithmetic for transform computations with a syntax element in an encoded bitstream, wherein the syntax element is operable to communicate the decision to a digital media decoder; and

outputting the encoded bitstream.

8. The digital media encoding method of claim 7 wherein said making a decision comprises:

verifying whether the lower precision arithmetic for transform computations produces a same decoder output as using a higher precision arithmetic for transform computations; and

deciding whether to use the lower precision arithmetic based upon said verifying.

9. The digital media encoding method of claim 7 wherein said lower precision arithmetic is a 16-bit arithmetic precision.

10. The digital media encoding method of claim 7 further comprising:

making a decision whether to apply a scaling of the input digital media data prior to transform coding; and

representing the decision whether to apply the scaling with a syntax element in the encoded bitstream.

11. The digital media encoding method of claim 10 wherein said making a decision whether to apply a scaling comprises deciding not to apply scaling of the input digital media data when losslessly encoding the digital media data.

12. A digital media decoding method comprising:

receiving a compressed digital media bitstream at a digital media decoder;

parsing a syntax element from the bitstream signaling choice of precision modes for transform computations during processing of the digital media data;

in the case that a first precision mode using scaling is signaled, scaling output of the decoder;

in the case that a second precision mode without scaling is signaled, omitting to apply scaling of the output; and

outputting a reconstructed image.

13. The digital media decoding method of claim 12 wherein said scaling output of the decoder comprises rounded division of the output by a number.

14. The digital media decoding method of claim 12 wherein said rounded division of the output is a rounded division by the number 8.

15. The digital media decoding method of claim 12 further comprising:

parsing a second syntax element from the bitstream signaling whether to use a lower arithmetic precision for transform computations during processing of the digital media data;

decoding blocks of transform coefficients from the compressed digital media bitstream; and

in the case that the second precision mode without scaling and the use of lower arithmetic precision are signaled, performing inverse transform processing of the transform coefficients using the lower arithmetic precision.

16. The digital media decoding method of claim 15 wherein said lower arithmetic precision is a 16-bit arithmetic precision.

17. The digital media decoding method of claim 12 wherein the digital media data is encoding using a two stage transform structure having a first stage transform followed by a second stage transform on DC coefficients of the first stage transform, the digital media decoding method further comprising:

decoding digital media data from the compressed digital media bitstream;

applying an inverse second stage transform to the digital media data;

applying an inverse first stage transform to the digital media data;

performing color conversion of the digital media data; and

wherein said scaling output of the decoder in the case that the first precision mode using scaling is signaled, comprises;

left shifting the digital media data by a single bit before input to the inverse first stage transform; and

right shifting the digital media data by three bits after the color conversion.

18. The digital media decoding method of claim 12 wherein said compressed digital media bitstream is encoded according to a syntax scheme defining separate primary image plane and alpha image plane for an image, the syntax element signaling choice of precision mode being signaled per image plane, whereby the precision mode of the primary image plane and the alpha image plane are independently signaled, and the decoding method comprises performing said actions of parsing the syntax element signaling choice of precision mode for each image plane, and in the case that the first precision mode using scaling is signaled for a respective image plane, scaling output of the decoder for the respective image plane.