Research of Image Mosaic Algorithm Based on Harris Corner Detection

This paper studies the general steps and methods of the image matching technology. Using a set of optimized Harris corner detection algorithm, a similar measure extracting feature points by the maximum correlation coefficient to achieve accurate matching. Finally, the weighted smoothing is performed to realize the seamless splicing of the two images according to the bilinear interpolation of the adjacent position of the overlap region. Introduction With the continuous development of imaging technology, image stitching technology is widely used in medical impression analysis, virtual reality and other fields. In this paper. Harris operator, detected corners an optimal set of eigenvalues, it is proved by experiments that the improved algorithm can effectively improve the extraction of feature points of the courier and accuracy, reducing the error rate of image matching. Harris corner and nature The corner point is also known as the point of interest, which is in every direction its neighborhood gradation change amount large enough. It is a kind of important image feature points, which contains rich one-dimensional structure information, and is widely used in various image processing techniques. In this paper, the selection of Harris conroner is as feature points for image matching. This operator has a matrix of the autocorrelation function associated M. The eigenvalues of the matrix M are first order curvature of the autocorrelation function. If its horizontal curvature and vertical curvature are higher than other points in the local neighborhood, it is considered that the point is the feature point. It is computationally simple and effective and is very stable. Its calculation is simple and effective and very stable, It is a feature extraction operator compared with other operators under the condition of image rotation, gray, noise and viewpoint transformation. Based on Harris corner theory, get the following corner extraction steps: (1) the use of horizontal, vertical difference operator for each pixel of image filtering in order to achieve x I and y I , then obtain four elements values of the matrix M: 2 2 2 2 , , x x y x x x y y y x y y I I I m I I I I I I I I I   = = × = ×       (2) The four elements of the matrix m are smoothed by Gauss filter,and the new matrix M is