Study Of Image Registration Technique And Application

Image registration, which is a fundamental problem in image processing, is a process to match two or more images of the same scene taken at different times, from different viewpoints, or by different sensors.The methods of image registration can be classified into two categories: the intensity-based matching approach and the feature-based matching approach. In this paper, the characteristics and the application fields of these two common methods are dissertated, and the feature-based matching approach is discussed as a main point.As one of the most common methods, the largest advantage of the feature-based approach is its ability of translating the analysis of the whole image into its features that contains the feature point, the feature curve, etc. And as a result, it speeds up the image process. The exactness of feature detection and orientation has influence on the matching process. In this paper, the current feature detection methods are discussed, and the common point-feature detection methods are emphasized. Finally, the one based on corner detection is adopted.After the precise and fast feature-point detection, the exact matching rule also affects the registration result. The disadvantage of common matching rules is that they cannot content the rotational invariance and ensure the exectness of detection. In this paper, a new image registration algorithm based on circles and Harris corner detection is described. This algorithm utilizes the rotational invariance of circle, and completes the registration combined with the mutual information. The simulation results show that the method can accomplish the registration for two rotated and translated images fleetly and stably.Finally, an important application field, scene mosaicing, is put forward. Based on the registration method in this paper, a new image fusion method is discussed. The experimental results prove that the method can significantly reduce the time of matching and solve the problem of joint.