Single Image Super Resolution Based On Multi-feature Fusion And Sparse Represetation

The image with high resolution can offer more details and is better to express the information. However, the image captured directly in practice may be low-quality?deform, blurry, down-sampled, noisy? due to the limitations of hardware device and the shooting environment. The direct approach to increase the resolution is to improve the image sensor manufacturing process, whereas the expensive cost will limit the application widely. Thus it is essential to improve the quality of image through the software method namely image super-resolution reconstruction technology.Instead of changing hardware equipment, image super-resolution reconstruction is to reconstruct a high resolution image from single or multiple low resolution images based on reasonable prior knowledge and mathematical model, which is widely used in many of image processing applications such as medical imaging, military, remote sensing and video surveillance. Compared with multi-image super-resolution, singleimage super-resolution utilizes only one low-resolution observation image, which has the wider applicable prospects. Therefore this research focuses on single image superresolution.This paper firstly reviews the domestic and foreign research status of superresolution reconstruction and several types of common classical algorithms that involves the basic principles and steps, advantages and disadvantages and so on. Among these methods, the single image super-resolution reconstruction technique based on sparse representation has been drawn attention of many scholars at home and abroad, has become a hot research in recent years. What's more, this method has been achieved the better effect relatively, so this paper carries on the study on the basis of this algorithm. The main contents is as follows:?1? The feature extraction is the important part in the super-resolution process, the effective feature representation of low-resolution image's high-frequency information can predict the corresponding high-resolution image accurately. Since the gradients extract the information only along the horizontal and vertical directions and the non-subsampled contourlet transform?NSCT? is poor relatively to capture the detailed information, to overcome the drawback, a novel super-resolution approach combined Gabor with NSCT is proposed to improve the quality of the image. The algorithm makes full use of the complementary of the Gabor transform and NSCT, to extract the texture feature using the Gabor transform and to extract the contour feature using the NSCT according to the characteristics of input image pieces, then perform the sparse coding reconstruction respectively, finally merge the pieces into a initial high-resolution image.?2?The input image is blurred more or less in the process of image degradation. However, if the input image is blurry, the sparse coding super-resolution may not be effective. To solve the problem, the approach revises the initial high-resolution image through the deblurred regularization l₁/ l₂ to eliminate the influence of blurred input. Finally, the experiments on various images demonstrate the effectiveness of the proposed algorithm.