Algorithm Research Of Image Processing And Registration Based On Features

Remote sensing technology is important to the development of the national economic development and military security needs. Remote-sensing image processing is a key step in remote sensing technology. As the center of remote sensing, image processing aspects of image registration studies in this area is always hot.First of all, this paper briefly introduces the field of image registration of the basic concepts and methods of geometric re-sampling .Geometric image registration as the final key step in the method determines the degree of match. Therefore it is necessary to have the correction and re-sampling theory, the application-specific to make adjustments for specific circumstances, on the whole to achieve a reasonable adjustment.Paper combines the needs of specific projects, based on wavelet transform features and multi-scale registration algorithm, and test the theory and application of the algorithm's stability and effectiveness. It provides a learning test image registration method, a measure of image similarity methods to effectively image registration evaluation. It will be a measure of image registration problem into a pattern classification problem, from design through the classification of machine learning to automatically test whether image registration, image registration test conducted. Tests show the feasibility of the method and reliability through programming.And the improved algorithm has the time and accuracy of contrast improved algorithm that can achieve requirements of the work, and proof of good stability of the algorithm.In this paper, the whole, to achieve the objectives of the research position, to achieve the requirements of the subject, the practical application of this project has been the experimental results. On the proposed registration to test the effects of the method to a wide range of effective registration Methods of test results, with the significance of further study.