Super-resolution Image Reconstruction Based SIFT On The Algorithm

Digital images are used widely, no matter what in the field of the military, medicine,surveying and mapping, or in the field of the security and detection. How to reconstruct aHigh-resolution image from the degraded images is a challenge. Super-resolution imagereconstruction uses a series of low-resolution images to reconstruct a single high resolutionimage. The purpose of the image registration is to estimate the geometry of the image movement.In this paper, we uses the scale invariant features registration method.The paper firstly introduced the generation, development and the research status of the superresolution image reconstruction. Then, this paper summarizes and analyzes commonly usedmethods of image super-resolution reconstruction, including the frequency domain analysismethod and the spatial analysis method. The spatial reconstruction methods are mainly includenon-uniform interpolation, filters, convex sets projection and maximum a posteriori probabilityFinally, the paper realize the convex set projection reconstruction algorithm with the scaleinvariant feature registration algorithm. The experimental data shows the algorithm put forwardin this paper is improved comparing to the traditional convex set projection algorithm from theMSE and the PSNR.The mainly work of the paper is:1. The paper proposes an adaptive parameter from the traditional convex set projectionalgorithm, and this effectively improves the convergence and the stability properties of thealgorithm.2. The paper realize the convex set projection reconstruction algorithm combined with thescale invariant feature registration algorithm. The experimental data indicates that the improvedalgorithm is improved comparing with the traditional algorithm in the MSE and the PSNR.