Variable localized bit-depth increase for fixed-point transforms in video coding

Patent metadata

Field	Value
Publication number	US-9185422-B2
Application number	US-201113159063-A
Country	US
Kind code	B2
Filing date	Jun 13, 2011
Priority date	Jul 15, 2010
Publication date	Nov 10, 2015
Grant date	Nov 10, 2015

Abstract

This disclosure describes techniques for mitigating rounding errors in a fixed-point transform associated with video coding by applying a variable localized bit-depth increase at the transform. More specifically, the techniques include selecting a constant value based on a size of a fixed-point transform in a video coding device and applying a variable localized bit-depth increase at the transform with a value equal to the constant value. Applying the variable localized bit-depth increase includes left-shifting a transform input signal by a number of bits equal to the constant value before the fixed-point transform, and right-shifting a transform output signal by a number of bits equal to the constant value after the fixed-point transform. The constant value is selected from a plurality of constant values stored on the video coding device. Each of the constant values is pre-calculated for one of a plurality of different transform sizes supported by the video coding.

First claim

The invention claimed is:

1. A method of coding a video signal, the method comprising:
storing, in a video coding device, a plurality of pre-calculated constant values, wherein each of the plurality of pre-calculated constant values indicates a variable localized bit-depth increase associated with one of a plurality of different transform sizes associated with video coding, and wherein each of the plurality of pre-calculated constant values is pre-calculated based on an input bit-depth value of the video signal, a transform bit-depth value associated with the one of the plurality of different transform sizes, and a transform bit-precision value associated with the one of the plurality of different transform sizes;
selecting a constant value from the plurality of pre-calculated constant values based on a size of a fixed-point transform in the video coding device; and
applying, at the fixed-point transform in the video coding device, the variable localized bit-depth increase indicated by the selected constant value by left-shifting a transform input signal by a number of bits equal to the constant value before the fixed-point transform, and right-shifting a transform output signal by a number of bits equal to the constant value after the fixed-point transform.

2. The method of claim 1 , wherein the fixed-point transform comprises a two dimensional fixed-point transform,
wherein selecting the constant value comprises selecting the constant value based on a size of one stage of the two dimensional fixed-point transform, and doubling the constant value; and
wherein applying a variable localized bit-depth increase comprises applying the variable localized bit-depth increase at the two dimensional fixed-point transform with a value less than or equal to the doubled constant value.

3. The method of claim 1 , the method further comprising:
determining whether an internal bit-depth increase is applied to the video signal; and
when the internal bit-depth increase is applied to the video signal, adjusting the constant value based on a value of the internal bit-depth increase, wherein applying a variable localized bit-depth increase comprises applying the variable localized bit-depth increase at the fixed-point transform with a value equal to the adjusted constant value.

4. The method of claim 3 , wherein adjusting the constant value comprises reducing the constant value, N, by the value of the internal bit-depth increase, B d , such that the adjusted constant value is equal to N−B d .

5. The method of claim 4 , wherein adjusting the constant value comprises reducing the constant value, N, by the value of the internal bit-depth increase, B d , when the value of the internal bit-depth increase is less than or equal to the constant value, such that (N−B d )≧0.

6. The method of claim 4 , wherein adjusting the constant value comprises reducing the constant value, N, to zero, when the value of the internal bit-depth increase, B d , is greater than the constant value, such that (N−B d )<0.

7. The method of claim 1 , further comprising applying an offset to the transform output signal prior to right-shifting the transform output signal by the number of bits equal to the constant value after the fixed-point transform.

8. The method of claim 7 , wherein the constant value is equal to N, and wherein applying an offset to the transform output signal comprises applying an offset equal to 2 N−1 to the transform output signal.

9. The method of claim 7 , further comprising combining the offset and right-shift for the variable localized bit-depth increase with other offsets and right-shifts applied to the transform output signal.

10. The method of claim 1 , wherein each of the plurality of pre-calculated constant values, N, is pre-calculated to satisfy B i +B TR N≦Z, wherein B i is the input bit-depth value of the video signal, B TR is the transform bit-depth value associated with the one of the plurality of different transform sizes, and Z is the transform bit-precision value associated with the one of the plurality of different transform sizes.

11. The method of claim 1 , wherein the plurality of different transform sizes comprises one or more of square transform sizes or rectangular transform sizes.

12. The method of claim 11 , wherein the plurality of different transform sizes comprises one or more of a 4×4 transform, a 4×8 transform, a 8×4 transform, a 8×8 transform, a 8×16 transform, a 16×8 transform, a 16×16 transform, a 16×32 transform, a 32×16 transform, a 32×32 transform, a 32×64 transform, a 64×32 transform, a 64×64 transform, a 64×128 transform, a 128×64 transform, or a 128×128 transform.

13. The method of claim 1 , the method being executable on a wireless communication device including the video coding device, the wireless communication device comprising:
a memory disposed within the video coding device to store the plurality of pre-calculated constant values;
a receiver configured to receive an encoded video signal;
a processor configured to decode the received encoded video signal, select the constant value from the plurality of pre-calculated constant values stored in the memory, and apply the variable localized bit-depth increase indicated by the selected constant value to the decoded video signal at the fixed-point transform in the video coding device.

14. The method of claim 13 , wherein the wireless communication device is a cellular telephone and the encoded video signal is received by the receiver and modulated according to a cellular communication standard.

15. The method of claim 1 , wherein the fixed-point transform in the video coding device comprises a fixed-point implementation of a discrete cosine transform.

16. The method of claim 15 , wherein the discrete cosine transform comprises a number of cascaded butterfly structures relative to the size of the discrete cosine transform.

17. A video coding device for coding a video signal, the video coding device comprising:
memory that stores a plurality of pre-calculated constant values, wherein each of the plurality of pre-calculated constant values indicates a variable localized bit-depth increase associated with one of a plurality of different transform sizes associated with video coding, and wherein each of the plurality of pre-calculated constant values is pre-calculated based on an input bit-depth value of the video signal, a transform bit-depth value associated with the one of the plurality of different transform sizes, and a transform bit-precision value associated with the one of the plurality of different transform sizes; and
a processor that selects a constant value from the plurality of pre-calculated constant values based on a size of a fixed-point transform in the video coding device, and applies, at the fixed-point transform, the variable localized bit-depth increase indicated by the selected constant value by left-shifting a transform input signal by a number of bits equal to the constant value before the fixed-point transform, and right-shifting a transform output signal by a number of bits equal to the constant value after the fixed-point transform.

18. The video coding device of claim 17 , wherein the video coding device comprises a video encoding device, the video encoding device further comprising an entropy encoding unit that encodes the video signal.

19. The video coding device of claim 17 , wherein the video coding device comprises a video decoding device, the video decoding device further comprising an entropy decoding unit that decodes the video signal.

Assignees

• Joshi Rajan L
• Chien Wei-Jung
• Karczewicz Marta
• Reznik Yuriy
• Chen Peisong
• Chuang Hsiao-Chiang
• Qualcomm Inc

Inventors

• Joshi Rajan L
• Chien Wei-Jung
• Karczewicz Marta
• Reznik Yuriy
• Chen Peisong
• Chuang Hsiao-Chiang

Classifications

• H04N19/60

  using transform coding · CPC title

• H04N19/61

  in combination with predictive coding · CPC title

• H04N19/157

  Assigned coding mode, i.e. the coding mode being predefined or preselected to be further used for selection of another element or parameter · CPC title

• H04N19/196

  being specially adapted for the computation of encoding parameters, e.g. by averaging previously computed encoding parameters (processing of motion vectors H04N19/513) · CPC title

• H04N19/46

  Embedding additional information in the video signal during the compression process (H04N19/517, H04N19/68, H04N19/70 take precedence) · CPC title