Motion estimation is an important part of the Moving pictures expert group-4 (MPEG-4) encoder, due to its significant impact on the bit rate and the output quality of the encoder sequence. Unfortunately this feature occupies a significant part of the encoding time especially when using the straightforward Full Search algorithm. For frame based video encoding, a lot of fast algorithms have been proposed, which have proved to be efficient in encoding In this paper We proposed a new algorithm named Cross based gradient descent search (CBGDS) algorithm, which is significantly faster than FS and gives similar quality of the output sequence. At the same time, We compare our algorithm with some other algorithms, such as Three step search (TSS), Improved three step search (ITSS), New three step search (NTSS), Four step search (4SS), Diamond search (DS), Block based gradient descent search (BBGDS) and Cellular search (CS). As the experimental results show, our algorithm has its advantage over the others. For objects based video encoding, most of the existing fast algorithms are not suitable because the arbitrarily shaped objects have more local minima. So we incorporate the alpha information and propose a new algorithm, which is compatible with the previously proposed efficient motion estimation method for arbitrarily shaped video objects.
展开▼
