Difference detection with encoder adaptability for low complexity surveillance video compression

Video surveillance has been widely applied in broad area to improve the public safety and to protect the residential privacy. Because of the requirements in storing the video data over a period of time for future use and because of the huge size of the raw video data, video compression technique has been integrated into the video camera of the surveillance system. As a state-of-the-art video compression technique, H.264/AVC is generally in use. However, it often leads to high computational complexity, which induces high power dissipation when a tremendous amount of cameras are installed to build a large scale of surveillance system. In this paper, a difference detection algorithm is proposed to work as a preprocessing module before video encoder to reduce the computational complexity of video compression. Because of the independency between the proposed processing and the video encoding scheme, it has high adaptability to work with any existing H.264 video encoder to save the cost in implementation and application. Experimental results demonstrate that an average of over 82% of computational complexity reduction can be achieved by integrating the proposed algorithm either into an encoder with high complexity rate distortion optimization (RDO) coding scheme or an encoder with fast algorithm. Besides the significant reduction in computational complexity, an average of 0.06 dB improvement in PSNR is further achieved against the H.264 Main Profile with high complexity RDO coding scheme.