Skip to Main Content
Entropy coding has been investigated for motion-compensated interframe (MC) prediction followed by two-dimensional discrete cosine transform (DCT) for prediction error. In particular, variable word length coding methods for motion vector and transform coefficients have been discussed assuming low bit rate such as 384 kbits/s for transmission of videoconference sequences. For motion vector information, it is advantageous to employ a one-dimensional code set common to both horizontal and vertical components of motion vectors. The code set can be obtained using a combined distribution of the two components. In order to encode transform coefficients, different methods are applied to significant and insignificant blocks. Run-length coding is adequate for representing clusters of insignificant blocks. In encoding transform coefficients in significant blocks, a zone coding method which encodes transform coefficients within a minimum area enclosing all nonzero coefficients is suitable. Simulation of video sequences shows that a combination of the coding methods described here can achieve high coding efficiency for videoconference sequences.