An Image Mosaics Research Based On Feature Points Matching

Image mosaic is the mutual between the overlapping portions of the image sequencespace matching alignment, the resampling fusion after forming a containing every imagesequence information, wide angle of view scene, complete, high clear new image. Imagemosaic technology allows the image acquisition device is more general, using common digitalcamera can be obtained to meet the requirements of the image. This technique is widely usedin computer vision, image processing, computer image, medical image analysis and virtualreality fields.This paper focuses on the feature point-based image stitching technology research on thistopic, focus on the various steps of the image mosaic, key theoretical knowledge of imagestitching, image acquisition, image preprocessing, image registration and image fusiontechniques are conducted in-depth study and discussion,mainly completed the followingwork:1. Learn the basic concepts of image stitching, research significance, as well as itsapplication field, and the current domestic and foreign research on image mosaic technologyis briefly introduced.2. Elaborated image stitching involved in the relevant theoretical knowledge and the keytechnologies used, and laid a good theoretical basis for the next job.3. Image acquisition, analysis of the problems of the acquisition process, and accordinglyneed to pay attention to when the acquisition; The image mosaic pre-pre-treatment, describesthe pretreatment process of geometric correction, edge detection and noise suppression of therole of several steps, as well as the methods used, and for these methods were compared, andfinally select the most suitable for the image of this pretreatment method. 4. Analyze and compare the principle of the algorithm used in the field of image stitching,as well as their advantages and disadvantages, the selected image registration method basedon features of the method be improved, and on this basis, an improved algorithm makesimagesmore accurate matching results; In order to improve the accuracy of the image featurematching, where we use the RANSAC algorithm, and this algorithm has been improved, thefeature points obtained by further purification, in order to be able to make image smoother.5. Finally, using the weighted average method of stitching after the image fusion, imagewith no visible seams, obtained better effects of splicing.