Image Super-resolution Study Of Dictionary-based Learning

With the rapid development of technology, images and more important in the transmission of information in humans. However, due to limitations in the imaging system and hardware natural conditions by, people often can not easily obtain a clear high-resolution images.Low-resolution images to high-resolution image have a very important value.In recent years the rise of a super-resolution image-based learning algorithm, this algorithm can not only be able to reconstruct a picture of a low-resolution image, high-resolution images, but without limiting multiple image to enlarge. This paper selection algorithm innovations in image reconstruction based super-resolution study, based on the proposed reconstruction algorithm based on a dictionary to learn new super-resolution image.Ideological line of this paper is to propose a MOD based dictionary learning new image super-resolution reconstruction algorithm on the basis of joint thinking on the dictionary, first extract a few blocks from a low-resolution image in the image, and the image blocks to get a pair of dictionary learning high and low resolution, dictionary, learning algorithm is used here, such as the optimal direction Engan (method of optimal directions, MOD) algorithm, which proposed to replace Lee Yang FFS algorithm used in the algorithm, and then use the dictionary to block low-resolution images sparse representation and reconstruction, reconstruction of the final results of further optimized to achieve better reconstruction results.Experimental is on related image and the other one is not related to image two issues.The result show that the proposed algorithm to select fewer training samples, dictionary training and faster reconstruction and reconstruction image with higher quality.