Adaptive-Rate Compressive Sensing for Monitoring Video Based on Fast Sparsity Estimation

Abstract

Theoretically, Compressive Sensing (CS) could sample and compress a signal when the whole signal is not captured and stored at the sampling side. However, it makes the estimation of signal sparsity difficult in the Adaptive-Rate Compressive Sensing (ARCS). In order to estimate the sparsity, a new ARCS method for monitoring video is proposed. The sparsity of the current frame of the video signal is estimated by observing the CS result of the previous frame, the computational complexity of sparsity estimation is simplified. Experiment results show that for each frame in the video, the proposed method can estimate its sparsity properly and achieve good reconstructed image quality. The proposed method reduces the requirement of sampling hardware, and makes it more practical in the field such as CS-based distributed video coding.