Fully-Convolutional Measurement Network for Compressive Sensing Image Reconstruction

Recently, deep learning methods have made a significant improvement in compressive sensing image reconstruction task. However, it still remains a problem of block effect which degrades the reconstruction results. In this paper, we propose a fully-convolutional network, where the full image is directly measured with a convolutional layer. Thanks to the overlapped convolutional measurement, the block effect is removed. In addition, because of the jointly training of the measurement and reconstruction stages, the adaptive measurement can be obtained. Furthermore, to enhance the performance of the network, residual learning is used in the reconstruction network. Experimental results show that the proposed method outperforms the existing methods in both PSNR and visual effect.