Assignment 5

1.       MPEG-4 motion compensation is supposed to be VOP-based. At the end, the VOP is still divided into macroblocks (interior macroblock, boundary macroblock, etc.) for motion compensation.

(a)     What are the potential problems of the current implementation? How would it compare to the current implementation?

(b)     Can there be true VOP-based motion compensation? How would it compare to the current implementation?

2.       Motion vectors can have subpixel precision. In particular, MPEG-4 allows quarterpixel precision in the luminance VOPs. Describe an algorithm that will realize this precision.

3.       What is the computational cost of SA-DCT, compared to ordinary DCT? Assume the video object is a 4×4 square in the middle of an 8×8 block.