Research And Implementation Of Fast Image Matching Algorithm

In the growing information society, the image matching technology has become a significant area of image processing research. Internet Image searchig engine technology, military aircraft target recognition technology, industrial robot vision technology, medical image processing techniques all use image matching techniques. Currently, researchers have proposed a variety of image matching algorithms. These matching algorithms needed to evaluate its unified index, a measure of the matching algorithm performance indicators include matching accuracy, matching real-time, matching probabilityIn this paper, HOG features as the feature descriptor, proposed image matching algorithm based on HOG features. After the algorithm for image segmentation, using the integral histogram method to quickly build Histogram of Oriented Gradients with HOG feature vectors to be matched to measure the similarity between the image and the template image,and obtain the best matching position. The algorithm can guarantee the accuracy of the match in the premise, as opposed to NCC algorithm greatly improved image matching algorithm in real time.In addition, based on the idea of space dimension reduction, this paper presents a fast matching algorithm. The algorithm projects conventional two-dimensional to a one-dimensional image space image space, and then extends the two-dimensional variable correlation coefficient formula to a one-dimensional variable coefficient, and use it as a similarity measure. The algorithm can not only maintain good of matching accuracy and matching probability, but also in real time on the NCC algorithm compared with the two orders of magnitude improvement.Finally, for the image matching algorithm performance evaluation, the MFC-based image matching algorithm software platform for fast matching algorithm is proposed in this paper, and other matching algorithm to do a comparative experiment and analyze the results.