Block compressed sensing based on human visual for image reconstruction

Abstract

Block Compressed Sensing (BCS) is one of the fundamental theories for image reconstruction. Compared with the traditional Compressed Sensing (CS) technique, it reduces the computational complexity and improves the efficiency of the reconstruction. However, the reconstruction quality of BCS is deteriorated to some degree. In order to improve the reconstruction quality of BCS, a new method based on human visual characteristics is proposed following the analysis of the DCT coefficients of an image. In the new method is introduced the contrast sensitivity in Watson visual model, indicating that the human eyes have different sensitivity to different DCT coefficients. To each element of the observation matrix in the same image block is assigned different weights based on visual characteristics. Finally, the experimental results demonstrate that the proposed approach can not only effectively improve the image reconstruction quality, but also have better subjective visual effect.