Method for computing motion vectors

摘要

The method involves converting motion vectors related to half images to vectors related to full images so that the vectors are exclusively related to the last image. Missing vectors of one or more full image blocks are detected by filtering vectors of blocks in the same or last image. Block line and column pairs with the lowest x-component and y-component scatter are selected from the processed vectors. The global vectors are computed for each full image using only selected lines and columns further selected from a medium scatter region.