Harris feature vector descriptor

This paper defines a new image feature called Harris feature vector, which is able to describe the image gradient distribution in an effective way. By computing the mean and the standard deviation of the Harris feature vector in a local image region, novel descriptors are constructed for feature matching which are invariable to image rigid transformation and linear intensity change. Experimental evidence suggests that the novel descriptor for point matching has a good adaptability to slight view point changing, JPEG compression and nonlinear changing of intensity, besides, the descriptor for line matching performs well too.