The Research Of Single Remote Sensing Image Super-resolution Reconstruction And Classification

Image super resolution reconstruction is a technique to reconstruct one or more high resolution images with low resolution images from different viewing angels, different time and different sensors of the same scenario. This paper concentrates on the high resolution reconstructing algorithm with single image and remote sensing image classification.In the progress of researching the single frame image super resolution reconstruction problem, a reconstructing method combining K means with local linear embedding algorithm. Since patches are similar with each other after blocking, it should be first clustered to different classes, extract one or more patches from different classes based on its similarity, then reconstruct the clustered images with local linear embedding algorithm. With such method, efficiency and effectiveness both are promoted. After experiments that reconstruct ordinary digital images and remote sensing images respectively applied with interpolation, local linear embedding, and combination of K means and local linear embedding algorithm, the method we proposed in this paper could results great conclusion. In the research of remote sensing image classification, a Fuzzy K means algorithm based on over-relaxation iteration is proposed. Introducing the over-relaxation iterative algorithm to the Fuzzy K means makes the convergence faster while without any accuracy. This algorithm is adaptive to large scale remote sensing image classification.