视觉传感器网络协作块压缩感知图像传输方法

Collaborative image transmission using block compressed sensing for visual sensor networks

JPEG2000等现有图像压缩传输算法虽然压缩性能好,但复杂度高,资源受限的视觉传感器节点很难适应。基于压缩感知理论研究视觉传感器网络环境下的图像传输问题。首先设计主-从节点对模型,分块监测图像;然后研究协作块压缩感知理论,基于图像分块和节点协作来减少测量矩阵的维数,降低压缩感知的计算复杂度;对处在发送端的主从节点,提出尽力而为的压缩感知测量值传输方法,以最大化图像重构质量为目标,尽力传送最多的测量值;而在接收端,汇聚节点依据内积准则来判断重构质量,确定传输终止条件,避免多余的测量值传输,节省传输耗能。实验结果表明,和经典压缩感知相比较,4个视觉传感器节点协作的块压缩感知传输方法将测量矩阵维数减少了2个量级,且有更少的网络耗能。

Most existing image compression algorithms,such as JPEG 2000,can not be applied to visual sensor networks with limited resources because of their high computational complexity.In this paper,the image transmission problem based on compressed sensing technique for visual sensor networks is addressed.First,a pair-wise sensor model is constructed,and the images are partitioned accordingly.Next,the collaborative block compressed sensing theory is studied.The dimensions of the measurement matrix are reduced via image blocking and collaboration among the nodes.The computational complexity and corresponding required computation cost are declined.Then,a best-effort transmission algorithm is proposed,in which the nodes transmit the measured data aiming at maximizing the reconstruction quality.For the fusion node at receiving end,an inner product criterion for evaluating the reconstruction quality is presented.Thus the redundant data transmission is avoided,and hence corresponding transmission energy is reduced.Experimental results show that,compared with traditional compressed sensing technique,the proposed algorithm implemented by four collaborative nodes reduces the measurement matrix by two orders of magnitude,and consumes less network energy.