Efficient motion-estimation algorithm for reduced frame-rate video transcoder

摘要: In multimedia applications, it is often necessary to adapt the bit rates of coded video streams available bandwidths various communication channels. Because different networks may have bandwidths, a gateway can include transcoder in order provide services users on networks. transcoding, motion estimation usually not performed because its heavy computational complexity. To speed up operation, reuses decoded vectors (MVs) from incoming bitstream. Previously, bilinear interpolation and forward dominant vector selection methods were proposed compose missing MVs for reuse. We propose new algorithm, called activity selection, which utilizes quantized discrete cosine transform coefficients residual blocks composing MV ones. addition, motion-vector refinement algorithm variable step-size search presented. The performance be improved while maintaining low complexity reduced frame-rate transcoder.