Research On Multi-View Matching Algorithm

Image matching is a hot topic in image processing and computer vision research.It is the research foundation of image restoration,image fusion and 3D reconstruction.The matching effect will directly affect the subsequent image processing.Multi-view matching is matched with multiple images obtained with different cameras at different perspectives.At present,the algorithm of image matching is divided into two categories:gray correlation and feature matching.The matching algorithm based on gray correlation makes full use of all the information of the image,The corresponding points in the image are judged by the correlation measure method through the pixel gray scale in the image,however this algorithm is slow because it is computationally expensive so it is not suitable for multi-view matching.In the image feature points matching research,the commonly used algorithms are:SIFT,ORB,SURF etc,in which SIFT algorithm remains unchanged not only to scale scaling,rotation,brightness changes,but also maintain a certain degree of stability on the viewing angle changes and the noise.Therefore,this paper selects the multi-view matching algorithm based on SIFT feature.The advantages of multiple images can be obtained by using an array camera at one time,the main work of this paper is as follows:(1)The acquired multi-view image is denoised enhancement,color calibration.(2)SIFT-based multi-view matching is performed on multiple sets of images that have been calibrated.(3)The multi view matching result is applied to the synthesis of high dynamic range images.The experimental results show that the advantage of the multi view matching algorithm in this paper is that the stability of the matching effect on the illumination and the change of the angle of view.The high dynamic range image synthesized by the multi-view matching algorithm is better than the Debevec algorithm,which is mainly embodied in the detailed information of the image and the expansion of the dynamic range of the image.This algorithm is expected to be applied in the practical application of the micro camera.