Earticle

With the study of human visual system (HVS), people find that the human eye shows different degree of sensitivity to different light signals. A higher background brightness region of the human eye for the distortion, the degree of sensitivity will be greatly reduced. The human eye’s sensitivity will be greatly reduced to the distortion of a higher brightness background region. This finding is discribed in Weber-Fechner law. Therefore, the encoder can be improved by employing this characteristic of human visual system. In this paper, the authors use this characteristic to improve the H.264 video coding algorithm. In order to improve the compression efficiency of encoder, give the brightness of different regions with different levels of quantification by adjusting the quantization step size (QP), without affecting the subjective quality of video. Experimental results show that the output bit rate decrease up to 10% -20% by using this improved algorithm in H.264 reference encoder, and the subjective quality of decoded image by using this improved algorithm is as good as the standard H.264 encoders.

Visual sensor networks (VSN) are wireless sensor networks in which each sensor has video capture and processing capability.Power consumption may be examined for encoding, transmitting, and receiving subsystems, and research has been performed on minimizing these power levels in parallel. When multiple camera modules of a visual sensor node are aimed at the same objects with different fields of view (FOVs), the captured images may overlap. Such overlapped FOVs give rise to encoding redundancy over the VSN and also lead to increased power consumption among adjacent nodes. The power-rate-distortion (P-R-D) is determined and used to construct an optimization problem for minimizing power consumption of each node, hence maximizing node lifetime. The optimal solution provides distributed power allocation and node scheduling over the VSN at the same time, via simple information sharing, resulting in network lifetime maximization.