Research Of Image Registration And Image Mosaic Algorithm

Image Mosaic technology is a very important research focus in computer vision.It is widely used in medical image processing, video detection, remote sensing, video compression and retrieval,3D virtual scene construction field and so on.In this paper,Image Registration Mosaic based on features and Image Fusion have been analyzed and researched separately.The main task are carried out as follows:1. Image Registration methods.This paper describes several common image registration methods, detailing the corner detection method based on the edge of the image and based on Harris corner detection.2. Image Registration based on SIFT image feature points.As the anti-noise performance of the method based on corner detection is poor, and does not have good invariance, this paper gives a image matching algorithm based on SIFT image feature points.The algorithm extracted SIFT feature points firstly, then use the nearest neighbor search algorithm for matching feature points, combined with RANSAC algorithm mismatching points.Experimental results show that the SIFT feature points have good translation invariance and rotational invariance.It also has better noise immunity performance compared to the Harris corner. At the same time,the experimental results show that the same image can be detected more SIFT feature points than Harris corner,so SIFT algorithm can achieve precise registration.3. Fast image matching based on SURF feature points. With the shortcomings of the large calculation amount and long time consuming of matching algorithm based on SIFT feature points, a fast algorithm based on SURF for image registration is presented in this paper. Firstly, exact feature points using SURF algorithm. For each feature point, the dominant orientation is assigned by computing Haar wavelet responses, and then the descriptors are generated. Then match the feature points using the optimized nearest neighbor search algorithm (BBF), and removed pseudo match points with RANSAC algorithm. At last, according to the actual need to select n most similar matchings. Experimental results show that this algorithm meets the requirements of accuracy with a small amount of calculation and fast speed advantages.4. Image Mosaic.Introduce three gray interpolation algotithm,and image mosaic realized using SIFT feature points matching.This paper also describes three fusion algorithms, weighted average method,pyramid image fusion and wavelet fusion.The data of each layers associated after image pyramid decomposition. So this kind of decomposition is redundant without directivity.The decomposition layers of wavelet are independent and have direction.So we can get a better visual fusion image using wavelet fusion.