Sub-block transform coding of prediction residuals
Techniques and tools for sub-block transform coding are described. For example, a video encoder adaptively switches between 8×8, 8×4, and 4×8 DCTs when encoding 8×8 prediction residual blocks; a corresponding video decoder switches between 8×8, 8×4, and 4×8 inverse DCTs during decoding. The video encoder may determine the transform sizes as well as switching levels (e.g., frame, macroblock, or block) in a closed loop evaluation of the different transform sizes and switching levels. The encoder and decoder may use different scan patterns for different transform sizes when scanning values from two-dimensional blocks into one-dimensional arrays, or vice versa. The encoder and decoder may use sub-block pattern codes to indicate the presence or absence of information for the sub-blocks of particular blocks.
1. A computer system comprising:
a buffer, implemented using memory, configured to receive encoded data in a bitstream, wherein the encoded data includes sub-block pattern information, the sub-block pattern information indicating an information pattern for presence or absence in the bitstream of prediction residual data for plural sub-blocks of a block of a video frame; and
a video decoder, implemented using one or more processors and memory, configured to perform operations to decode the encoded data to reconstruct the video frame using a variable-block-size inverse frequency transform, wherein the operations include, for a given sub-block of the respective sub-blocks of the block:
using the sub-block pattern information to determine that the given sub-block has prediction residual data in the bitstream;
getting, from the encoded data in the bitstream, encoded transform coefficients for the given sub-block;
entropy decoding the encoded transform coefficients for the given sub-block;
inverse quantizing the transform coefficients for the given sub-block;
with the variable-block-size inverse frequency transform, converting the transform coefficients for the given sub-block to prediction residual values for the given sub-block; and
combining the prediction residual values for the given sub-block with predicted values for the given sub-block.
2. The computer system of claim 1 , wherein the block is an 8×8 block and the plural sub-blocks are 4×4 sub-blocks.
3. The computer system of claim 1 , wherein the variable-block-size inverse frequency transform uses a transform size selected from among plural available transform sizes.
4. The computer system of claim 3 , wherein the plural available transform sizes include 8×8 and 4×4.
5. The computer system of claim 1 , wherein, for each of the respective sub-blocks of the block, if the sub-block has at least one non-zero transform coefficient, the sub-block pattern information indicates presence in the bitstream of prediction residual data for that sub-block.
6. The computer system of claim 1 , wherein the operations further include:
selecting a scan pattern from among multiple scan patterns; and
using the selected scan pattern to reorder the transform coefficients for the given sub-block from a one-dimensional array to a two-dimensional array.
7. The computer system of claim 1 , wherein the operations further include:
performing motion-compensated prediction to determine predicted values for the block, the predicted values for the block including the predicted values for the given sub-block.
8. One or more computer-readable media having stored thereon encoded data in a bitstream, the encoded data including sub-block pattern information, the sub-block pattern information indicating an information pattern for presence or absence in the bitstream of prediction residual data for plural sub-blocks of a block of a video frame, the one or more computer-readable media being selected from the group consisting of non-volatile memory, optical media, and magnetic media, wherein the encoded data is formatted to facilitate reconstruction by operations comprising:
receiving the encoded data in the bitstream; and
decoding the encoded data to reconstruct the video frame using a variable-block-size inverse frequency transform, wherein the decoding includes, for a given sub-block of the respective sub-blocks of the block:
using the sub-block pattern information to determine that the given sub-block has prediction residual data in the bitstream;
getting, from the encoded data in the bitstream, encoded transform coefficients for the given sub-block;
entropy decoding the encoded transform coefficients for the given sub-block;
inverse quantizing the transform coefficients for the given sub-block;
with the variable-block-size inverse frequency transform, converting the transform coefficients for the given sub-block to prediction residual values for the given sub-block; and
combining the prediction residual values for the given sub-block with predicted values for the given sub-block.
9. The one or more computer-readable media of claim 8 , wherein the block is an 8×8 block and the plural sub-blocks are 4×4 sub-blocks.
10. The one or more computer-readable media of claim 8 , wherein the variable-block-size inverse frequency transform uses a transform size selected from among plural available transform sizes.
11. The one or more computer-readable media of claim 10 , wherein the plural available transform sizes include 8×8 and 4×4.
12. The one or more computer-readable media of claim 8 , wherein, for each of the respective sub-blocks of the block, if the sub-block has at least one non-zero transform coefficient, the sub-block pattern information indicates presence in the bitstream of prediction residual data for that sub-block.
13. The one or more computer-readable media of claim 8 , wherein the decoding further includes:
selecting a scan pattern from among multiple scan patterns; and
using the selected scan pattern to reorder the transform coefficients for the given sub-block from a one-dimensional array to a two-dimensional array.
14. The one or more computer-readable media of claim 8 , wherein the decoding further includes:
performing motion-compensated prediction to determine predicted values for the block, the predicted values for the block including the predicted values for the given sub-block.
15. In a computer system that implements a video encoder, a method comprising:
encoding a video frame, to produce encoded data in a bitstream, using a variable-block-size frequency transform with support for switching of transform size at varying levels within the video frame, wherein the encoding includes, for a given sub-block of plural sub-blocks of a block of the video frame:
using predicted values for the given sub-block to determine prediction residual values for the given sub-block;
with the variable-block-size frequency transform, converting the prediction residual values for the given sub-block to transform coefficients for the given sub-block; and
quantizing the transform coefficients for the given sub-block;
entropy coding the quantized transform coefficients for the given sub-block to produce encoded transform coefficients for the given sub-block; and
setting sub-block pattern information to indicate that the given sub-block has prediction residual data in the bitstream; and
outputting the encoded data in the bitstream, wherein the encoded data includes the encoded transform coefficients for the given sub-block and the sub-block pattern information, the sub-block pattern information indicating an information pattern for presence or absence in the bitstream of prediction residual data for the plural sub-blocks of the block of the video frame.
16. The method of claim 15 , wherein the block is an 8×8 block and the plural sub-blocks are 4×4 sub-blocks.
17. The method of claim 15 , wherein the variable-block-size frequency transform uses a transform size selected from among plural available transform sizes.
18. The method of claim 17 , wherein the plural available transform sizes include 8×8 and 4×4.
19. The method of claim 15 , wherein, for each of the respective sub-blocks of the block, if the sub-block has at least one non-zero transform coefficient, the sub-block pattern information indicates presence in the bitstream of prediction residual data for that sub-block.
20. The method of claim 15 , wherein the encoding further includes:
selecting a scan pattern from among multiple scan patterns; and
using the selected scan pattern to reorder the transform coefficients for the given sub-block from a two-dimensional array to a one-dimensional array.