Research For Image Mosaic Based On Feature Points

As time being,the demand for high-resolution images is becoming more and more urgent.Image mosaic technology has also become a popular research field in digital image processing.It is widely used in virtual reality,intelligent transportation,medical imaging and other areas.The most important constraints in image mosaic are the accuracy of image registration and the time comsum of the algorithm.In this paper,an improved RANSAC algorithm based on conditional constraint model is proposed,which is aiming at the problem that the registration accuracy is low and the time comsum is high.For the poor integration,this paper proposes an improved progressive fade fusion algorithm based on the Gaussian model,which eliminates the problem that the traditional progressive fade fusion method is prone to splicing gap,and achieves better fusion effect.The main work of this paper is as follows:Firstly,the feature point detection algorithms based on SIFT and SURF are introduced in details.The nearest neighbor feature matching algorithms based on k-d tree and BBF are introduced.The performance of the feature point detection algorithm and the feature point matching algorithm are analyzed by a large number of experiments.The indexes include the number of feature points and matching points,the matching rate and the running time.Secondly,for the traditional RANSAC algorithm exist a huge data set,too many false matches and a lot of iterations,a new RANSAC algorithm based on conditional constraint model is proposed in this paper.This algorithm establishes a conditional constraint model for the original matching point set,eliminates the points that donot satisfy the constraint condition,improves the proportion of the interior points,and reduces the number of iterations of the RANSAC algorithm and increases the efficiency of the algorithm.Finally,the problems that the splicing gaps and the transition are not smooth in the traditional image fusion algorithm,this paper studies a progressive fade image fusion algorithm based on the Gaussian model,which supresses the splicing gaps,and achieves a better fusion result.In this paper,a complete algorithm was coded in VC++ development environment and the Opencv library,and a large amount of experionmental images were processed.The experimental results show that the improved RANSAC algorithm based on the conditional constraint model can effectively reduce the time comsum of the algorithm.At the same time,the fusion algorithm based on the Gaussian model can remove the splicing gaps in the image and obtain a smooth fusion result.