Region Adaptive Measurements for Distributed Compressive Video Sensing

Abstract

Abstract. Due to limited resources, wireless video sensor networks (WVSN) needs low complexity methods to realize video capture and compression. Recently, distributed video coding (DVC) has emerged to reduce the video encoding complexity with a certain coding efficiency. Furthermore, compressive sensing (CS) has been proposed to directly capture the compressed data efficiently. In this paper, combining DVC and CS, a novel video coding framework has been designed for WVSN. In order to preserve low complexity at the encoder, odd frames of the original video can be compressed as key frames by standard intra-coding while even frames can be processed as CS frames by CS encoder. Here, flexible mode selection of the encoder is applied to improve coding efficiency. Furthermore, adaptive measurements are allocated for different blocks in view of the object and background regions in the video frames. At the decoder, the bi-directional dictionary is proposed as the sparse basis to improve the recovery quality of CS. The experimental results validate the effectiveness of the proposed scheme with better performance than other compared schemes.