Super-resolution Restoration Methods Based On Contourlet Transform Of Image

With the development of modern technology, higher image quality is required in more aspects of life. In order to meet the requirement of application, people tend to work on certain algorithms with the purpose of achieving high quality image instead of improving the device with high costs. Research on image super-resolution recon- struction (SRR), which is an image-enhancing approach aiming to reconstruct a high-resolution (HR) image from observed multiple low-resolution (LR) images, has been done recently. Nowadays, SRR techniques have been widely used in many aspects of life.As an efficient set of techniques in signal processing, wavelet transform enjoys good local features in both time domain and frequency domain of signal representation. Accordingly, the multi-scale analysis tools, which enjoy better quality, have been developed. Scholars have already put forth algorithms on wavelet-based SRR, and have achieved good performance in signal construction. Inspired by the ideas above, this paper introduces the multi-scale analysis tools into SRR technique, and more specifically describing, investigates the SRR algorithms based on Contourlet transform.The main work related includes the contents as follows:(1) A SSR method for single-frame image is proposed, which is based on Contourlet transform combined with Soft-decision adaptive interpolation (SAI) algorithm. The algorithm takes advantage of edge and detail capture features via Contourlet transform and details preserving ability in enhancement via SAI. Experimental simulation indicates that the proposed algorithm boosts quite good performance. And then, based on the analysis of the method above, an improved algorithm is proposed, with feasibility proof via experimental results.(2) A SSR method for sequential images is proposed in the paper, which is based on Contourlet transform combined with Projection Onto Convex Set (POCS) algorithm. POCS algorithm is classical in sequential image reconstruction techniques, whose advantage is that the prior information of images could be easily added and the edges and details of HR-images could be well preserved as well. Compared with other algorithms, it is shown via experiment that the proposed algorithm boosts both better subjective visual quality and higher PSNR values.(3) A SSR method for sequential image is proposed in the paper, which is based on Contourlet transform combined with image fusion technique. Image fusion is to synthesize images of a special scene captured by two or more sensors, and generate a new description of the scene, with the aim to reduce the uncertainty of images and improve the definition through information complement of several images. Based on it, the proposed algorithm is finished with Contourlet transform. Compared with other algorithms, experimental simulation shows its rationality and feasibility.