Image Super-resolution Based on Compressive Sensing

Based on Compressive Sensing, we introduce sparse signal representation theory to modify the local geometric similarity model and construct sparse geometric similarity representation. Based on the modified model we can estimate the optimized reconstruct coefficients by jointing the original global and local image structure themselves, without the support of other training image database. The experimental results show that the algorithm can greatly improve the reconstruction of the edge and texture details in the high-resolution image.