Spatiotemporal Remote Sensing Image Fusion Base On Spares Representation

In order to obtain high resolution remote sensing image,in addition to improve the hardware facilities, people put forward the idea of the low resolution image reconstruction to fuse the high resolution image.In this paper, the method of using sparse representation and difference image of low resolution image fusion, get high resolution images. Main contents are the following three points:Sparse Representation method as a kind of way for the image expression, there are a lot of applications. We can use the method of sparse expression and super-resolution reconstruction theory, to reconstruct the existing low spatial resolution remote sensing image, to obtain high spatial resolution remote sensing image. In the theory of sparse representation, K- SVD method because of its simple, fast and flexible characteristics is widely used. The advantage of the characteristics of image signal,taking remote sensing image into the signal,then using super complete dictionary training method for remote sensing image expression. Getting rid of the noise, image representation, compression, the method has good effect and significance.Reference in the Modis image and Landsat image based low- high imaging method for reconstructing high resolution image(Huihui song,Bo Huang,2013). Obtained by the method of reconstruction of high resolution image. But this method is not very good in a wide range of terrain changes got very good reconstructed image. For this, This paper puts forward an improved method to join difference image feature. Using to join the difference image,it’s very good to get the change of features. When in remote sensing image is reconstructed,we can relate the difference image and reconfiguration of remote sensing image. The remote sensing images obtained from the K- SVD method has the characteristics of difference image(namely,feature change characteristics of the different time).It is better solved the original method in a wide range of terrain changes.Using down sampling Landsat 8 remote sensing image and GF satellite remote sensing image to the improved fusion methods for implementation and verification. Through the image contrast with the result of the experiment, The improved fusion method than the original fusion method in a wide range of features changes, has a good experimental effect. But the improved fusion methods are still insufficient, For example, when we want to reconstruct a wide range of water surface,the improved fusion method can’t get a very good result. It needs to be future improved, it can be more close to the real remote sensing images.