Research Of Image Matching Algorithm

Image matching refers to the use of effective matching algorithm for a given two ormore same or similar image search key process. Image matching is a very importantproblem in digital image processing, and also is the basis of many computer vision theoryand application. In medical image processing and analysis, remote sensing monitoring,industrial automatic, weapon guidance, and machine vision applications, the imagematching technology are indispensable key step. Image with strong structural features, suchas corners, edges, statistics and texture features, most of these features in the imagematching technology plays an important role, many key depends largely on thecharacteristics of image matching problem of selecting, testing and expression. Selectionfor different image matching problem, due to the characteristics of the different, matchingresults may have bigger difference.Image matching technology based on the participation degree of the people can bedivided into semi-automatic match and fully automatic match. Semi-automatic matchingbased on human-computer interaction way to extract features (such as corner), then usingcomputer image features matching, transform and resampling. Automatic matching iscompletely based on human, using computer directly to the program can stay as long as thematching of image input, the other process does not require the participation of people.This thesis mainly studies image matching issues which based on the characteristics,summarizes the common methods of image preprocessing, and presents concrete examples,and the common image matching methods are summarized. Image matching method basedon the characteristics usually has the following steps: first step is the necessary imagepreprocessing, enables the images to more convenient and better work for the followingservices; second step is to match two images or multiple image feature extraction, featurespace; third step is to extract in the feature space of the feature matching and find thereference diagram features in corresponding to match the picture and complete thematching position. According to the above steps, in this paper, on the basis of summarizingthe predecessors’ research, this paper proposes a new feature matching algorithm based onHarris corner points, the algorithm and combining with angular point around the grayscaleinformation of image matching. Image preprocessing of to match first, and then carried outon the reference graph and matching graph respectively Harris corner detection and recordtheir coordinates, to detect the Angle of point grey value and its grey value proportion within the neighborhood, and then set a grayscale threshold and threshold percentage. Isconsidered for matching the image feature points and corresponding feature points in thereference image gray level difference and accounts for the proportion of whether within theset threshold, has the characteristics of high precision, high speed.