The Improved Image Mosaicking Algorithm Based On Harris Corner Feature Matching

With the development of the digital image processing techniques, the demand of people is becoming more and more, also more and more high, the application of panoramic image is no exception. Panoramic image can be obtained by two ways, one is through specialized hardware equipment; another is achieved by image mosaicking. The price of former is higher, but the image quality is higher; the latter cost is very low, but the picture quality is poorer. Image mosaicking technology is mainly to improve the quality of image under the premise of the low cost. Image mosaicking technology focus on image registration and image fusion technology, and research of most scholars focus on the optimization of image registration. The process of image mosaicking is mainly composed of image preprocessing, image registration and image fusion. This paper makes a deep research on image registration.In this paper, the improved image mosaicking algorithm based on Harris corner detection was proposed. First, the auto-adaptive threshold theory was adopted to generate more accurate corners. Compared with the traditional algorithm, the improved algorithm greatly reduced the number of false corners, while increasing the number of correct corner, and reducing the amount of data. To describe the corner neighborhood window, Hu moments were adopted. The spurious feature points were rejected by RANSAC algorithm and build affine transformation model between the images, and calculate model parameters, to achieve image registration. Finally, the weighted average algorithm was used to fusing image, in order to achieve image mosaicking.