Efficient interframe transform coding using temporal context

Abstract

Three dimensional (3D) transform coding can reduce the interframe redundancy among a number of consecutive frames, while the motion compensation technique can only reduce the redundancy of at most two frames. The former is very efficient when the correlation between interframe pixels is high. However, the performance will be degraded for complex scenes with a large amount of motions. This paper presents a three dimensional discrete cosine transform (3D-DCT) coding with variable temporal lengths, which is based on the local temporal activity. Two scene change detectors are used to detect the local temporal activity. Our idea is to let the motion activity in each block (with variable temporal length) be very low, while the efficiency of the 3D-DCT coding is increased. Through intensive computer simulations, the performance of the proposed 3D-DCT coding has been found to have substantial improvement over the conventional fixed length 3D-DCT coding. Furthermore, it is significant to point out that the performance of our proposed algorithm is better than that of the MPEG coding.