Sub-block transform coding of prediction residuals

In a computer system, a computer-implemented method of video encoding, the method comprising:

• selecting a transform size from among plural available transform sizes for one or more blocks of prediction residual data representing differences between pixel values of a motion-compensated prediction block and pixel values of a corresponding block of video, wherein the motion-compensated prediction block has a motion-compensated prediction block size;
• and
• encoding the one or more blocks of prediction residual data, wherein the encoding includes processing the one or more blocks of prediction residual data with one or more frequency transforms of the selected transform size;
• and
• signaling transform size switching information in an encoded video bitstream, wherein the transform size switching information specifies the selected transform size but not the motion-compensated prediction block size.

A computer-readable medium storing computer-executable instructions for causing a computer system programmed thereby to perform a method comprising:

• evaluating plural available transform sizes for motion-compensated prediction residual data representing differences between pixel values of a motion-compensated prediction block and pixel values of a corresponding block of video, wherein the motion-compensated prediction block has a motion-compensated prediction block size;
• selecting one or more of the plural available transform sizes based at least in part upon results of the evaluating;
• encoding the motion-compensated prediction residual data using frequency transforms of the selected one or more transform sizes;
• and
• signaling transform size switching information in an encoded video bitstream, wherein the transform size switching information indicates the selected one or more transform sizes but not the motion-compensated prediction block size.

In a computer system, a computer-implemented method comprising:

• receiving transform size switching information signaled in an encoded video bitstream;
• selecting a transform size from among plural available transform sizes for prediction residual data representing differences between pixel values of a motion-compensated prediction block and pixel values of a corresponding block of video, wherein the motion-compensated prediction block has a motion-compensated prediction block size, the received transform size switching information indicating the selected transform size for the prediction residual data without indicating the motion-compensated prediction block size;
• and
• decoding the prediction residual data, wherein the decoding includes processing the prediction residual data with one or more inverse frequency transforms of the selected transform size.

In a computer system, a computer-implemented method comprising:

• receiving transform size switching information signaled in an encoded video bitstream;
• selecting a transform size from among plural available transform sizes for one or more blocks of prediction residual data representing differences between pixel values of a motion-compensated prediction block and pixel values of a corresponding block of video, wherein the motion-compensated prediction block has a motion-compensated prediction block size, wherein the received transform size switching information specifies the selected transform size but not the motion-compensated prediction block size;
• and
• decoding the one or more blocks of prediction residual data, wherein the decoding includes processing the one or more blocks of prediction residual data with one or more inverse frequency transforms of the selected transform size.