Research On The Fast And High Precision Matching Algorithm For Image Super-resolution Reconstruction

Image reconstruction is an important technology in the field of computer visionresearch, which is a process that puts multiple images with displacement andcomplementary information into a higher resolution single image. Firstly, the technologyobtains multiple images for the same target; secondly, it uses the image registrationalgorithm for calculating the displacement of each image caused by camera shake; finally,a higher resolution single image can be obtained by the reconstruction algorithm. Becausethe technology can significantly improve the image quality, Image reconstruction has agood application prospect in many fields, such as video monitoring, medical diagnosis,remote sensing, astronomical observation, biological information identification and so on.In order to meet the speed of image reconstruction in practical application, theresearch work in this paper focuses on fast registration algorithm, including featureextraction, feature point positioning, feature matching algorithm and implementation basedon OpenCV platform.In this paper, the main contribution can be summarized as follows:1. According to the low speed of feature extraction using traditional algorithms. Withan analysis of the feature extraction algorithms and strategies, a fast feature extractionalgorithm is proposed, and its advantage is high speed and precision, and it can meet therequirements of quasi real-time image reconstruction.2. The paper studies the theory of Harris corner location and locates the corner frompixel level to sub-pixel level and then the pyramid Lucas-Kanade algorithm is used tosolve the problem of large and incoherent motion, Finally the RANSAC algorithm isapplied to remove false matching points, and it not only reduces the probability of mismatch but also improves the matching accuracy.3. After comprehensive analysis and comparison with all kinds of reconstructionalgorithms’performance, this paper uses wiener filtering for image restoration.4. Based on OpenCV platform, the paper completes the experiment. It includes featureextraction, feature matching and image reconstruction.Finally, the summary and the trend of the research on image registration would begiven at the last chapter.