Compressed video bitstream is sensitive to errors that may degrade the reconstructed images severely even the bit error rate is small. One approach to combat the impact of error is error concealment at the decoder without increasing the bit rate and changing the encoder. We propose motion vector based error concealment algorithms to recover the motion vector per pixel instead of that per block according to the relation of neighboring motion vectors. The displacement per pixel can be estimated more accurately by using the tendency of neighboring motion vectors. Besides, we use not only the relation among motion vectors, but also the pixels. The pixels of the error block are divided into different parts according to their consistency with neighboring blocks and the displacement at each pixel of these parts is interpolated by relative motion vectors. From simulation results, the proposed motion vector based methods provide better reconstruction quality for damaged images than other methods.

The three main reasons why the new H.264 (MPEG-4 AVC) video coding standard has a significant performance better than the other standards are the adoption of variable block sizes, multiple reference frames, and the consideration of rate distortion optimization within the codec. However, these features incur a considerable increase in encoder complexity. As for the multiple reference frames motion estimation, the increased computation is in proportion to the number of searched reference frames. In this paper, a fast multi-frame selection method is proposed for H.264 video coding. The proposed scheme can efficiently determine the best reference frame from the allowed five reference frames. Simulation results show that the speed of the proposed method is over two times faster than that of the original scheme adopted in JVT reference software JM73 while keeping the similar video quality and bit-rate.

When the frame size is downscaled for video transcoding, the new motion vector (MV) must be computed. This paper presents an algorithm to utilize the activity measurement by DC value and the number of non-zero quantized DCT coefficients in the residual macroblock to compose the motion vector. It can reduce the complexity for motion estimation and improve the performance of the spatial domain video transcoder.