Research On Methods For Repairing Missing Information On Images

As a type of information symbol, images have great superiority over texts and become more and more important mode of information expression. Images are also the most important data source in Geographic Information Science and image processing is a very crucial step in remote sensing application. However, data missing phenomena are very common in our daily lives and science research, such as damaged images and occlusion. How to repair the damaged part and fill some information into the occluded regions becomes an attention of some people. At present, there are two themes closely related to this topic, one is image inpainting, and the other is texture synthesis. How to restore the missing information on natural images and the remote sensing images is the focus of this dissertation. In order to obtain this purpose, we carry out some research on the principle, methods and applications about image inpainting and texture synthesis.The grey value of image can be treated as regionalized variable, which has the property of both random and spatial structure distribution. We introduce the ideology and viewpoints of geostatistics into image inpainting. Considering the property of anisotropic distribution of local area on the image, we presented a novel anisotropic interpolation model. This image inpainting method overcomes the traditional methods and obtains fairly good result.We discuss the principle and methods of texture synthesis. Aiming at some disadvantages of the actual algorithms in texture synthesis, we present some new algorithms and new solutions. We introduce the markov random field model, which is used in texture synthesis, into image inpainting and carry out some research on new algorithms based on wavelet decomposition.Information missing is not only a common phenomenon in our daily lives, but also occurs in science research. We carry out special analysis about the mechanism of occlusion on the remote sensing image, and discuss the advantages and disadvantages of occlusion as well as summarize some countermeasures at present. Next, we introduce the notion of true-orthophoto, the principle and method of true-orthophoto generation and discuss the causation of information missing. The difficulties concerning information missing are analyzed. Combining with the particularity of occlusion compensation on the remote sensing image, we present some cognitions and...