Image Compressive Sensing via Multiple Constraints

Total variation (TV) based method is widely used in image compressive sensing recently, which also achieves great success in under-sampled recovery due to its virtue for edges preserving. However, since TV favours piecewise constant solution, it shows bad performance in fine structure recovery. To overcome this shortcoming, the paper proposed a novel image compressive sensing scheme by introducing the balanced sparsity model into the conventional TV based compressive sensing. The proposed scheme takes full advantage of the multiple constraints, which constrain the different priors in true image respectively, and benefits from both of them to improve the recovery quality. Experimental results demonstrate that the novel scheme outperforms the compared compressive sensing methods in both of objective indicator and visualization.