Coaxial Different Field Picture Image Registration Algorithm

Image registration is a basis in the field of image processing tasks, and it is an extremely important part. Most image processing tasks rely heavily on image registration. Image registration is a necessary pre-processing of these image processing tasks. Quality of its result is an enormous impact on the entire image processing. Some image processing tasks-such as image stitching-can not do without the support of image registration. If the image has not been correct registering, image stitching cannot carry on.The main task of this paper is an image registration between a part of a picture based on global image as the background and the global image itself. This aim is to make the observer can not only get the basic information of the global image and can also track and monitor the details which are interested. In addition, the information of the uninterested regional in the image can be weakened. This is to reduce the attention of the observers on irrelevant information, and also to reduce the amount of image data. This makes the subsequent image processing faster.For the goals set above, this paper studies on a two picture image registration. This paper researches on two pictures which are filmed in the same direction but different focal length and field, and register them. The way of the image registration based on the feature point. Due to the filmed of two pictures, the orientation of the camera has not changed, but the focal length changed. The shape of the objects in the image is similar, and only the size of the objects has changed. So the positional relationship of the two graphs can be estimated from the two corresponding objects.This paper respectively discusses the image feature extraction algorithm, feature point matching algorithm and image transformation parameters estimated. And it gives out some image processing results to compare. According to the actual results, it gives out analytics and error discusses. This paper detailedly discusses the feature point extraction algorithm. It mainly uses the Susan-corner-detecting algorithm. And it compares the differences between the Susan-corner-detecting algorithm and other algorithms. Through experiments, it shows that the Susan-corner-detection algorithm has a good performance of corner detection.