Low-complexity sign prediction for video coding

1. A method of coding video data, the method comprising:

identifying a set of pre-stored residual samples in a pre-stored portion of the video data;

assigning a first absolute value to a first transform coefficient of each pre-stored residual sample of the set of pre-stored residual samples;

assigning a second absolute value to all remaining transform coefficients of each pre-stored residual sample of the set of pre-stored residual samples except for the respective first transform coefficients that are assigned the first absolute value, the second absolute value being different from the first absolute value;

applying an inverse transform to each pre-stored residual sample of the set of pre-stored residual samples to form a set of dequantized reconstructed residual samples, wherein each pre-stored residual sample of the set of pre-stored residual samples comprises a dequantized transform coefficient block;

combining the set of pre-stored residual samples to form a combination of pre-stored residual samples;

applying a first sign combination to transform coefficients of the combination of pre-stored residual samples to obtain a first hypothesis reconstruction with respect to a current block of the video data;

applying a second sign combination to the transform coefficients of the combination of pre-stored residual samples to obtain a second hypothesis reconstruction with respect to the current block of the video data;

deriving respective cost functions with respect to the first hypothesis reconstruction and the second hypothesis reconstruction;

comparing the respective derived cost functions;

selecting, based on the comparison, either the first hypothesis reconstruction or the second hypothesis reconstruction; and

coding the current block using a sign prediction associated with the selected one of the first hypothesis reconstruction or the second hypothesis reconstruction.

2. The method of claim 1 , wherein the first absolute value is one (1), and wherein the second absolute value is zero (0).

3. The method of claim 1 , further comprising:

scaling each dequantized residual sample of the set of dequantized residual samples using a dequantized value associated with each of the first transform coefficients; and

reconstructing the current block using the set of scaled dequantized residual samples.

4. The method of claim 1 , wherein one or more pre-stored residual samples of the identified set of pre-stored residual samples are stored at a fractional precision in the stored portion of the video data.

5. The method of claim 1 , further comprising scaling each pre-stored residual sample of the identified set of pre-stored residual samples using a pre-defined factor.

6. The method of claim 1 , further comprising clipping each pre-stored residual sample of the identified set of pre-stored residual samples using a pre-defined factor such that each clipped pre-stored residual sample can be represented using four (4) bits or less of data.

7. The method of claim 1 , wherein combining the set of pre-stored residual samples to form the combination of pre-stored residual samples comprises performing a linear combination of the set of pre-stored residual samples to form a linear combination of pre-stored residual samples.

8. The method of claim 7 , wherein performing the linear combination of the set of pre-stored residual samples comprises calculating a weighted sum of the set of pre-stored residual samples, at least in part by assigning a respective weight to each pre-stored residual sample of the set of pre-stored residual samples.

9. The method of claim 1 , wherein combining the set of pre-stored residual samples to form the combination of pre-stored residual samples comprises merging the set of pre-stored residual samples to form a single merged residual block, the method further comprising:

overwriting the set of pre-stored residual samples with the single merged residual block in a memory device.

10. The method of claim 1 , further comprising:

deriving a single residual block using one or more pre-stored residual samples of the identified set of pre-stored residual samples

overwriting the set of pre-stored residual samples with the derived single residual block in a memory device.

11. The method of claim 1 , wherein the stored portion of the video data including the set of pre-stored residual samples is stored to a first memory device, the method further comprising:

loading one or more pre-stored residual samples of the set of pre-stored residual samples from the first memory device to a second memory device on one of a once-per picture basis or a once-per largest coding unit (LCU) basis.

12. The method of claim 1 , wherein coding the video data comprises decoding encoded video data, wherein the current block is an encoded block of the encoded video data, and wherein coding the current block comprises reconstructing the encoded block.

13. A device for coding video data, the device comprising:

a video data memory; and

processing circuitry coupled to the video data memory, the processing circuitry being configured to:

identify a set of pre-stored residual samples in a portion of the video data that is pre-stored to the video data memory;

assign a first absolute value to a first transform coefficient of each pre-stored residual sample of the set of pre-stored residual samples;

assign a second absolute value to all remaining transform coefficients of each pre-stored residual sample of the set of pre-stored residual samples except for the respective first transform coefficients that are assigned the first absolute value, the second absolute value being different from the first absolute value;

apply an inverse transform to each pre-stored residual sample of the set of pre-stored residual samples to form a set of dequantized reconstructed residual samples, wherein each pre-stored residual sample of the set of pre-stored residual samples comprises a dequantized transform coefficient block;

combine the set of pre-stored residual samples to form a combination of pre-stored residual samples;

apply a first sign combination to transform coefficients of the combination of pre-stored residual samples to obtain a first hypothesis reconstruction with respect to a current block of the video data;

apply a second sign combination to the transform coefficients of the combination of pre-stored residual samples to obtain a second hypothesis reconstruction with respect to the current block of the video data;

derive respective cost functions with respect to the first hypothesis reconstruction and the second hypothesis reconstruction;

compare the respective derived cost functions;

select, based on the comparison, either the first hypothesis reconstruction or the second hypothesis reconstruction; and

code the current block using a sign prediction associated with the selected one of the first hypothesis reconstruction or the second hypothesis reconstruction.

14. The device of claim 13 , wherein the first absolute value is one (1), and wherein the second absolute value is zero (0).

15. The device of claim 13 , wherein the processing circuitry is further configured to:

scale each dequantized residual sample of the set of dequantized residual samples using a dequantized value associated with each of the first transform coefficients; and

reconstruct the current block using the set of scaled dequantized residual samples.

16. The device of claim 13 , wherein one or more pre-stored residual samples of the identified set of pre-stored residual samples are stored at a fractional precision in the portion of the video data stored to the video data memory.

17. The device of claim 13 , wherein the processing circuitry is further configured to scale each pre-stored residual sample of the identified set of pre-stored residual samples using a pre-defined factor.

18. The device of claim 13 , wherein the processing circuitry is further configured to clip each pre-stored residual sample of the identified set of pre-stored residual samples using a pre-defined factor such that each clipped pre-stored residual sample can be represented using four (4) bits or less of data.

19. The device of claim 13 , wherein to combine the set of pre-stored residual samples to form the combination of pre-stored residual samples, the processing circuitry is configured to perform a linear combination of the set of pre-stored residual samples to form a linear combination of residual samples pre-stored to the video data memory.

20. The device of claim 19 , wherein to perform the linear combination of the set of pre-stored residual samples, the processing circuitry is configured to calculate a weighted sum of the set of pre-stored residual samples, at least in part by assigning a respective weight to each pre-stored residual sample of the set of pre-stored residual samples.

21. The device of claim 13 , wherein to combine the set of pre-stored residual samples to form the combination of pre-stored residual samples, the processing circuitry is configured to merge the set of pre-stored residual samples to form a single merged residual block, and wherein the processing circuitry is further configured to overwrite the set of pre-stored residual samples with the single merged residual block in the video data memory.

22. The device of claim 13 , wherein the processing circuitry is further configured to:

derive a single residual block using one or more pre-stored residual samples of the identified set of pre-stored residual samples

overwrite the set of pre-stored residual samples with the derived single residual block in the video data memory.

23. The device of claim 13 ,

wherein to code the video data, the processing circuitry is configured to decode encoded video data,

wherein the current block is an encoded block of the encoded video data, and

wherein to code the current block, the processing circuitry is configured to reconstruct the encoded block.

24. An apparatus for coding video data, the apparatus comprising:

means for identifying a set of pre-stored residual samples in a pre-stored portion of the video data;

means for assigning a first absolute value to a first transform coefficient of each pre-stored residual sample of the set of pre-stored residual samples;

means for assigning a second absolute value to all remaining transform coefficients of each pre-stored residual sample of the set of pre-stored residual samples except for the respective first transform coefficients that are assigned the first absolute value, the second absolute value being different from the first absolute value;

means for applying an inverse transform to each pre-stored residual sample of the set of pre-stored residual samples to form a set of dequantized reconstructed residual samples, wherein each pre-stored residual sample of the set of pre-stored residual samples comprises a dequantized transform coefficient block;

means for combining the set of pre-stored residual samples to form a combination of pre-stored residual samples;

means for applying a first sign combination to transform coefficients of the combination of pre-stored residual samples to obtain a first hypothesis reconstruction with respect to a current block of the video data;

means for applying a second sign combination to the transform coefficients of the combination of pre-stored residual samples to obtain a second hypothesis reconstruction with respect to the current block of the video data;

means for deriving respective cost functions with respect to the first hypothesis reconstruction and the second hypothesis reconstruction;

means for comparing the respective derived cost functions;

means for selecting, based on the comparison, either the first hypothesis reconstruction or the second hypothesis reconstruction; and

means for coding the current block using a sign prediction associated with the selected one of the first hypothesis reconstruction or the second hypothesis reconstruction.