Research On Moving Object Ghosting Method Based On SVM In Panoramic Mosaic Image

In recent years, the rapid development of digital image technology has made the application of image Mosaic technology become a very popular technology, it flexibly applied in aerospace, military applications, remote sensing images, virtual reality, medical image analysis, human-computer interaction, computer vision, and other fields. Because image Mosaic technology can splice two or more than two images which have overlap area seamless into a panoramic image, avoiding the limitations of existing imaging equipment, making only relying on ordinary camera can be taken out of the big Angle of high-resolution images.Nowadays, research of image Mosaic has become a hot research in the field of image processing. However, acquisition of image sequence will disturb by moving targets inevitable, if without any processing, the formed panoramic image often appear problems like virtual shadow and obvious flat-fell seam.In this paper, we study the panoramic image Mosaic with moving targets, combined with the theory of statistics, committed to eliminate the inaccurate matching of feature points and the unnatural of image fusion, a seamless Mosaic image method is proposed. To eliminate the influence of the moving targets, through the combination of support vector machine (SVM) method to extract the moving object. Then, using the scale invariant feature points extraction, classifying feature points, and updating and transforming model parameters, to obtain precise transformation model. Finally, based on considering the color, texture, and its unique "movement implies", referring to regional growth figure cut method, we find the optimal spell juncture, finally get the panoramic images meet the requirements of high quality. The experimental results show that the method can effectively eliminate the virtual shadow phenomenon, weaken the splicing gap, and then get the seamless panoramic image Mosaic.