An adaptive motion vector composition algorithm for frame skipping video transcoding

Frame skipping transcoding involves periodically dropping certain incoming frames to reduce the bit rate of the outgoing stream, thereby causing many motion vectors extracted from the incoming stream to be invalid. Instead of re-calculating these invalid motion vectors, a number of techniques recompose the vectors by using the incoming information to reduce computational complexity. Reducing computational complexity is the key to real time operation. In this paper we propose an enhanced parametric activity dominant vector selection (PADVS) algorithm that reduces the overall computational complexity significantly, while retaining and even improving image quality.