Adaptive Compressive Sensing Based on Sparsity Order Estimation for Wireless Image Sensor Networks

For related applications based on wireless image sensor networks (WISNs), the efficiency of image signal acquisition and transmission is crucial. How to effectively reduce the data volume and the sampling rate during the signal communication process is an urgent research problem to be solved. Block compressive sensing (BCS) can be used to solve the above problem. In natural images, the sparsity order of different blocks tends to vary, indicating varying levels of complexity across these blocks. Therefore, it is crucial to set the appropriate sampling rate for each block. However, we cannot directly obtain the pixel domain information of the images on the sampling side. To solve the above problem, a novel adaptive rate CS scheme based on sparsity order estimation is proposed. We first estimate the variance of each block in the measurement domain according to the Johnson–Lindenstrauss lemma. Then find the complex block and the sparse block according to the variance values. An a priori probability model is used to estimate the sparsity orders of the complex block and the sparse block, respectively. In order to reduce the complexity of the sparsity order estimation process, the sparsity orders of the remaining blocks are estimated according to a simple signal cosine similarity estimation model. Finally, the sparsity order of each block can be used to calculate its corresponding sampling rate. Experiment results indicate that the proposed scheme can obtain higher reconstruction accuracy than state-of-the-art adaptive CS schemes. Moreover, the reconstructed images exhibit a significant improvement in visual quality.