Confusion on motion estimation block diagram-MPEG video compression-:

I am assuming, without reading the whole article, that frame # n-1 is an I frame, and frame # n is a P frame in that picture. Coding P frames involves two main steps:
  1. Find the best approximation with of the P frame using (moved) blocks from the I frame.
  2. Compute the difference between the approximated and actual P frames, then block-encode (using quantized DCT and entropy coding) that difference.
I.e., instead of coding frame # n first try to represent it as a "collage" of chunks (blocks) from frame # n-1, then encode the error of that representation.
 
Top