Research On Image Registration Technology Based On Point Feature

Image registration technology is a basic problem of image processing. It is the process that under different times, different sensors such as imaging equipment, or different conditions such as weather, illumination, camera position and angle two or more same scene images make registration. The purpose of image registration is to eliminate the geometric distortion of images and make the same scene of different images align or matching. Now, image registration technology is widely used in many fields such as aerial remote sensing images, medical diagnostics and so on. My study is on the image registration technology between two images.The algorithms of extracting and matching point feature are two key algorithms of image registration technology based on point feature.In this paper,we divide point feature into two categories as providing point feature by user and providing point feature by algorithm such as harris corner feature extraction algorithm.Firstly, we provided image registration algorithm based on providing point feature by users and the related simulation of this algorithm. we use these artificial provision matching feature points to get the space transform structure. Then we do space transforming and grayscale interpolation on unregistered image,and we achieved an initial align result.Secondly, againsting the result of image registration algorithm based on providing point feature by users mostly depends on the accuracy of the artificial selection matching feature point,we give the algorithm of image registration based on harris corner feature.we use the improved extracting harris corner feature algorithm provided by this paper to extract feature points. We use the way of Grayscale Cross-Correlation to match corner feature.Then we do some simulation analysis and achieve certain result.Finally, againsting the issue of the rotation sensitive on the way of Grayscale Cross-Correlation to match harris corner feature, we propose the improved algorithm of image registration based on harris corner feature. We improve the algorithm of harris corner feature matching. Basing on the way of Grayscale Cross-Correlation to obtain coarse matching harris corner feature,we use euclidean distance ratio to make harris corner feature match accurate and obtain good image registration results. Then we compare and analyze these three types of image registration algorithms. In the end,we propose the least squares method to make harris corner feature match accurate.And this way need simulation to prove it.