Subband Coding Algorithms for Video Applications: Videophone to HDTV-Conferencing

Various subband coding schemes are developed for high compression video coding applications. These schemes are based on two distinct interframe subband models. Both models re compared and, in terms of performance, shown to be equal under certain conditions. In the first model the subband decomposition is performed on the motion compensated prediction error signal and Is therefore located inside the DPCM loop. In the second model each frame of video signal is first split into narrow bands before each band Is coded by Interframe prediction. The first model, which Is less complex in terms of hardware Implementation, is considered for low speed video coding applications. In the second model however, the motion estimation and compensation prediction are performed at a reduced speed. It is, therefore, found to be suitable for higher quality video applications. In addition this model, due to its multilayer video representation, can be made compatible with existing codecs such as the recently proposed CCITT coder. In the coding process the motion compensated prediction error signals are entropy coded after being quantized. Initially, 1-dimensional (1-D) runlength coding is considered to transmit the positions of the nonzero quantized values. To further enhance the coding performance, an adaptive block runlength coding is proposed. The coding takes into consideration the position of each Individual band In the 2-dimensional (2-D) frequency spectrum of the original signal. Simulations were performed for a wide range of video quality applications; from videophone to HDTV video-conferencing. Subsequently, the results for transmission rates of 64 kb/s for videopbone and 5.625 Mb/s for HDTV-conferencing were presented.