Video compression can reduce the amount of data information in multimedia sensor networks. Traditional distributed video coding scheme that adopts regional unified coding may induce the distortion problem in decoding estimation of the intense motion region. This paper proposes an improved algorithm for DVC. The algorithm is based on image gradient field. The regions of the Wyner-Ziv frame that cannot be coded or decoded accurately are extracted according to the criteria of ROI and compressed with entropy coding. Other regions are encoded and decoded based on LDPC. The optimization of video transmission is achieved eventually. Simulation results show that: the proposed algorithm can enhance the encoding efficiency of the intense motion region; the algorithm can also reduce the rate and improve the quality of decoded image

and decrease the energy consumption of sensor nodes ultimately.