A multi-scale block compression sensing sampling algorithm

Abstract

All sub-blocks in the same layer use the same sampling rate, which leads to a deviation between the sampling and observation information in the multi-scale block compression sensing algorithm and the original image information, and results in reducing the accuracy of image reconstruction. Therefore a multi-scale block compression sensing algorithm (HEI) is proposed to reduce the deviation of information. Image blocks are decomposed to obtain multi-scale blocks in this method. And the high frequency coefficients of wavelet transform are used to extract the edge information, which is used as the adaptive sampling rate of sub-blocks. Experiments are carried out on 6 standard images, and the effects of different hierarchies of wavelet transform and the scale of graph subblocks on HEI method are analyzed. The experimental results show that, compared with the adaptive sampling method using low frequency information, the proposed method has better quality and visual effect on the reconstructed images of the test images.