Low-complexity wavelet-based image/video coding for home-use and remote surveillance

The availability of inexpensive cameras enables alternative applications beyond personal video communication. For example, surveillance of rooms and home premises is such an alternative application, which can be extended with remote viewing on hand-held battery-powered consumer devices. Scalable wavelet image/video coding is attractive for this application since the video can be scaled easily to many different formats including long-term viewing at low-power or even battery-operated devices. To this end, we present a modification to the SPECK wavelet image encoding algorithm which significantly improves efficient mapping on embedded systems. The key to our improvement is that the quadtree for bit assignment is stored in an efficient way such that the position of the first significant bit can be computed, so that later the code can be developed with minimal and direct memory access. This efficient storage is possible by structuring the data of the quadtree into nodes covering groups of pixels. Our approach gives a substantial reduction of the memory access and thus power consumption, which is essential for embedded applications and mobile devices.