Study On Image Registration Techniques With High Registration Ratio And Quick Velocity

Image registration is an important task of image processing. It is a basic step for image mosaic, object identification, image fusion, detection of changing object, analysis of sequential images etc. Application of image registration is involved in military affairs, remote sensing, medicine and computer vision etc. Now, there are lots of registration methods for multi-source image and data. Accordingly, image registration methods are divided by various rules. The common image registration methods are divided into two types: intensity-based methods and feature-based methods. Because of the disadvantage of lower velocity for intensity-based methods and lower registration ratio for feature-based methods, the paper makes researches for intensity-based methods with higher velocity and feature-based methods with higher registration ratio. Then, the paper makes attempt to synthesize these two methods, and expect to reach image registration with quick velocity and higher registration ratio. For the first type of methods, because of the disadvantage of lower registration velocity when searching more pixels and bigger area, the author presents the conception of "image trace" and gives some methods such as image registration algorithm based on image trace and based on threshold image trace. Moreover, the author present pyramid registration algorithm and two thresholds registration algorithm by changing the searching strategy. 中北大学学位论文For the other type of methods, because of the disadvantage of difficult feature choosing, complex feature extraction algorithms and lower registration ratio, the paper makes research for image registration based on wavelet decomposition, on projection and on entropy. Furthermore, the author give a new image feature—fractal dimension and apply it to image registration. For the synthesizing methods of two types of methods, the paper synthesizes the advantage of high registration ratio for intensity-based methods and quick velocity for feature-based methods and presents the image registration algorithms based on synthesizing method of image segmentation and contrast method, image wavelet and image trace, and improving fractal dimension. At last, the author make experimentation for the algorithms in the paper by various images from different sensors and finish the image registration system with higher registration ratio and velocity on the software system Visual C++ 6.0.