Image Super-resolution Reconstruction Algorithms Based On To Fuzzy

Scientific research and economic activities need for High Resolution (HR) images for precise cognition of image details. Restricted by the high price of imaging equipment, limitation for transmission and storage, and other factors, HR images are not easy to be popularized. Super Resolution methods can obtain HR images with existing imaging equipment and technologies by integrating information from Low Resolution Images (LRs) of the same imaging field. Although most traditional SR methods base on the global translation and rotation imaging model, local motion/blur frequently appears in LRs. So, it’s necessary to explore the use of the existing GLOBAL translational model based SR methods in LOCAL degradation scenarios.Based on the aforementioned fact, PSF models for motion blur and out-of-focus blur are introduced, and the estimation methods for corresponding parameters are analyzed and tested and results are showed; Classic Image Restoration algorithms are introduced for PSF based models. Then, SR related information is introduced:the information cognition ideology, imaging model, classic SR methods, challenges and opportunities for SR. At last, based on local motion/blur model, on the foundation of MAP SR algorithm and ROI, region de-blurring based SR algorithm is analyzed and introduced. The possibility for motion region extraction is analyzed (considered as prior knowledge for experiments in this thesis), and ROI model for local motion/blur region is elaborated. Potential problems for the use of ROI in SR are analyzed and separate/joint SR for both background and local blur region are realized. Improvement for motion estimation is made by integrating the advancement of region-based and feature-based motion estimation method. Experiment results show that region deblurring based algorithm is superior to translational SR algorithm at motion region. It’s a nice try to overcome the constriction of translation PSF model.