Estimating sample-domain distortion in the transform domain with rounding compensation

• US 8,249,145 B2
• Filed: 09/29/2011
• Issued: 08/21/2012
• Est. Priority Date: 04/07/2006
• Status: Expired due to Fees

• Current Assignee: Microsoft Technology Licensing LLC (Microsoft Corporation)
• Original Assignee: Microsoft Corporation
• Inventors: Chang, Cheng, Lin, Chih-Lung B., Regunathan, Shankar
• Primary Examiner(s): NGUYEN, LEON VIET Q
• Application Number:
  US13/248,784
• Publication Number:
  US 20120020409A1
• Time in Patent Office: 327 Days
• Field of Search: 375/240.01, 375/240.03, 375/240.16, 375/240.18, 382/239, 382/244, 382/250
• US Class Current: 375/240.01
• CPC Class Codes: H04N 19/126 Details of normalisation or weighting functions, e.g. normalisation matrices or variable uniform quantisers
• CPC Class Codes: H04N 19/61 in combination with predictive coding
• CPC Class Codes: H04N 19/625 using discrete cosine transform [DCT]

Abstract

Techniques and tools are described for compensating for rounding when estimating sample-domain distortion in the transform domain. For example, a video encoder estimates pixel-domain distortion in the transform domain for a block of transform coefficients after compensating for rounding in the DC coefficient of the block. In this way, the video encoder improves the accuracy of pixel-domain distortion estimation but retains the computational advantages of performing the estimation in the transform domain. Rounding compensation includes, for example, looking up an index (from a de-quantized transform coefficient) in a rounding offset table to determine a rounding offset, then adjusting the coefficient by the offset. Other techniques and tools described herein are directed to creating rounding offset tables and encoders that make encoding decisions after considering rounding effects that occur after an inverse frequency transform on de-quantized transform coefficient values.

Claims

1. An encoder comprising:

a frequency transformer for applying a frequency transform to convert samples in a sample domain into transform coefficients in a transform domain;

a quantizer for quantizing the transform coefficients;

an entropy encoder for entropy encoding the quantized transform coefficients;

an inverse quantizer for de-quantizing the quantized transform coefficients;

a rounding compensation module for adding a rounding offset to the de-quantized transform coefficients; and

a controller for making encoding decisions after considering post-inverse frequency transform rounding effects on de-quantized transform coefficient values, wherein the controller includes a module for estimating pixel-domain distortion between original transform coefficients of a block and corresponding de-quantized transform coefficients of the block in the transform domain after the rounding compensation module adds the rounding offset by compensating for rounding in at least one of the corresponding de-quantized transform coefficients of the block and computing the estimated pixel-domain distortion using the original transform coefficients of the block, the at least one rounding-compensated coefficient and other coefficients of the de-quantized transform coefficients of the block.

2. The encoder of claim 1 further comprising a table for use in the compensating by, for each of the at least one rounding-compensated coefficient:

determining an index from the coefficient;

looking up the index in the table to determine a corresponding offset; and

adjusting the coefficient by the corresponding offset.

3. The encoder of claim 1 wherein the encoding decisions are based at least in part on the estimated pixel-domain distortion.

4. The encoder of claim 1 wherein the encoding decisions comprise setting coding parameters for two or more of intra/inter encoding, quantization dead zone size, transform size, number of motion vectors, motion vector value, perceptual quantization, quantization thresholding, and quantization step size.

5. The encoder of claim 1 wherein the rounding effects indicate multiple values in a group of the de-quantized transform coefficient values affect distortion equivalently, and wherein the encoding decisions comprise selecting between the multiple values based on bits required for the multiple values respectively.

6. A method comprising:

identifying a range of values for a de-quantized transform coefficient;

computing a rounding offset for each of plural values in the range;

identifying a periodic pattern in the computed rounding offsets;

mapping representative values to corresponding rounding offsets in an offset table, wherein the corresponding rounding offsets show at least one period of the pattern without the offset table including all values in the range; and

storing the offset table.

7. The method of claim 6 further comprising distributing the offset table with a video encoder.

8. The method of claim 6, wherein the offset table maps representative de-quantized values to representative offsets for less than all of the de-quantized values.

9. The method of claim 6 further including:

determining an index from a first coefficient;

determining an offset by looking up the index in the offset table; and

adjusting the first coefficient by the offset.

10. A method comprising:

identifying a range of values for a de-quantized transform coefficient;

computing a rounding offset for plural values in the range;

identifying a periodic pattern in the computed rounding offsets; and

mapping representative values to corresponding rounding offsets in an offset table, wherein the corresponding rounding offsets show at least one period of the periodic pattern without the offset table including all values in the range.

11. The method of claim 10, further comprising distributing the offset table with a video encoder and storing the offset table.

12. The method of claim 10, wherein the offset table maps representative de-quantized values to representative offsets for less than all of the de-quantized values.

13. The method of claim 10, further including:

determining an index from a first coefficient;

determining an offset by looking up the index in the offset table; and

adjusting the first coefficient by the offset.