Research On Image Stitching Technology Based On Improved ORB Algorithm

Image stitching is an important computer vision technique that can fuse multiple images to achieve an extended field of view,improved image resolution,and information enrichment.This paper focuses on the two key steps in image mosaic,namely,image registration and image fusion technology.In the aspect of image registration,the concept of region division and similarity algorithm are used to extract regions with high similarity,and then the ORB algorithm is used to detect feature points for regions with high similarity to improve the efficiency of image feature point detection;secondly,the Hamming distance is used for rough matching;finally,the improved RANSAC algorithm is used to further remove the mismatched feature point pairs,and the partitioning sampling method and candidate model detection strategy iteratively to obtain the best data model which improve the accuracy of feature point registration.The algorithm in this paper is compared with SIFT,SURF and traditional ORB algorithms,and the results show that the algorithm has the advantages of high matching quality and fast speed under the image scaling,illumination and blurring phenomena.In terms of image fusion,an improved algorithm combining the best suture algorithm and weighted fusion algorithm is proposed to address the issues of unclear fusion region and stitching gaps in the existing algorithms.This algorithm firstly introduces the alignment difference and color difference when constructing the energy function of the best suture algorithm to enhance the distinction between aligned and unaligned positions in the overlapping regions;secondly,the improved best suture algorithm is used to fuse the suture region and the gradual in and out weighted fusion method is used to fuse the non-suture region to obtain the final fused image.The fusion results are comprehensively evaluated by subjective perception and objective indexes.The results show that the algorithm of this paper can make the image achieve smooth transition,solve the obvious stitching and ghost phenomenon,and improve the stitching quality.Finally,this paper designs an image stitching system based on the improved algorithm,and uses Qt framework to develop the system interface.Each functional component is integrated to realize image stitching and parameter display by the interface.