Bi-level Frame for Reduced Complexity in Video Compression

Communication between end to end is facilitated by the recent development in the area of wired and wireless networks. The end to end devices like mobiles, handheld and palm size PC’s have limited capability for computation, display capacity and bandwidth. This augment, for ways for communicating large data files like video, in highly compressed and acceptable quality over the limited available bandwidth. We propose a reduced complexity quad tree based recursive motion estimation algorithm. It is best suited for video transmission over a very low bandwidth network. The scheme involves the pre-processed quad tree decomposition of frame in bi-level domain, providing a better level of adaptation to scene content compared to fixed or variable block size approach in gray domain. The partition criterion is based on segmenting the blocks recursively into active or passive regions in bi level domain. Only the active regions (object) blocks, are partined to lower dimension (8*8 & /4*4). The active region is very small region compared to total frame size thus we are able to reduce the number of blocks for motion estimation and encoding. Our result shows that it reduces the prediction error, computational complexity and bit rate reduces significantly at considerable PSNR.