A point matching algorithm for brain CT images based on SIFT and gray feature

Abstract

On the basis of the original SIFT algorithm, this paper presents an improved point matching algorithm for brain CT image, which combines the SIFT and gray feature. Firstly, the SIFT algorithm is used to extract the key points of the images, and for each key points, a 128 dimensional SIFT feature vector is generated. The Euclidean distance of the SIFT feature vector is used as the similarity measure to obtain the initial matching point-pairs. Then, the gray method is used to extract the gray feature for each key point in initial matching point-pairs. Both the Euclidean distance and cosine similarity of gray feature vectors are used as the similarity measure to obtain the final matching point-pairs. This algorithm makes full use of the principle of rotation invariance and the gray level difference of key points to match. Experimental results show that this method improve the accuracy from 89.83% to 92.78% with the time complexity basically unchanged, at the same time, the number of correct matches are not much reduced.