A super-resolution algorithm based on adaptive sparse representation

To improve the performance of super-resolution reconstruction of images, a super-resolution algorithm based on adaptive sparse representation is proposed. Our algorithm regards the difference between the high-resolution image and the reconstructed image with Iterative back-projection algorithm as the image´s high-frequency characteristic, which is further used for high-resolution dictionary training. And after edge detection, our algorithm adaptively applies sparse representation and Iterative back-projection to edge patches and smooth patches respectively for reconstruction. Experimental results show that, with our algorithm the reconstructed image edges, especially the strong edges, are close to the original high-resolution image, and PSNR could be improved significantly.