Reserch In Iamge Super-Resolution Based On Sparse Throry

Image super-resolution technique is to recover a high-resolution image from one or more low-resolution images. This paper focuses on the single image super-resolution algorithm based on sparse theory and guided filter. Sparse representation (SR) is proposed as a new trend development of traditional signal representation theory and an extremely powerful tool. Recently, D. Donoho et al. proposed compressive sensing (CS), and successfully makes SR a breakthrough in practical applications. This paper applies image guided filter to extract low-resolution features when constructing dictionaries.Image super-resolution as a kind of technique at the bottom level of images is to recover a high-resolution clear image from one low-resolution fuzzy image. It also provides a better image quality environment for image analysis and image processing. Therefore, the image super-resolution technique can be used in many fields of the image processing and also has a huge potential development.The main research of this paper focuses on the problem of the single image super-resolution technique based on sparse theory and guided filter in this paper. The goal of this paper is to recover a high-resolution clear image from one low-resolution fuzzy image, and applies to face recognition tasks. The main contributions are as follows:(1) This paper briefly introduces the background and significance of current image super-resolution, sparse theory and face recognition.(2) This paper introduces the image super-resolution method based on sparse theory, image guided filter for extracting features and over-complete dictionary learning.(3) The extensive experimental results of comparing various traditional single image super-resolution with the single image super-resolution technique based on sparse theory and guided filter are given in this paper, which demonstrates the efficiency of our method.(4) The integration of the single image super-resolution technique with face recognition system. The experiments on some data sets show that the recognition rate could be enhanced when applying our proposed method to the input low resolution face images.