The Research Of Panoramic Images Mosaic Based On Image Sequences

The background of this paper is unmanned automatic identification system development. On the study of the image mosaic technology results, this paper summary the research work of the image pre-processing, image registration and image fusion. And primarily study the image registration technology.This paper comparative analysis the template-based image registration algorithm and the Harris corner feature-based image registration algorithm. But the two methods both have some limitation. For the above problem, using the classical SIFT algorithm extract the image feature points. In generating the feature vector description, using the normalized gradient algorithm improve the speed and accuracy of the feature matching algorithm. The use of RANSAC algorithm is better excluding the mistake matching points. Experiments prove that this algorithm is better than original in the feature matching time, the feature matching accuracy and the feature matching efficiency.In the analysis of the existing image mosaic and display methods, using the forward-mapping method obtain the re-sampled image. Then map the spliced images to the reference image, complete the image mosaic processing. Eliminate the image mosaic seams by using the gradually fade image fusion algorithm. Finally, process the entire sequence of images according to this method, obtain the panoramic image.According to the above description, this paper presents the analysis results. Experiments show that the algorithm in this paper has achieved more desirable results than previous in image registration and image fusion. It can be applied to the occasions of sequence image mosaic effectively.