Improving SURF Descriptors Performance for Image Matching

Image matching is one of the important issues in the machine vision science and is the basis of many systems in the field of artificial intelligence. It is mainly used in object recognition in robotic applications, data acquisition and analyzing the received data from satellite, object tracking in the military applications and image retrieval. The aim of this paper is to introduce an improved scale and rotation invariant algorithm based on the local feature for image matching demonstrating good performance in the case of intensity and blur changes. One of the important algorithms for image matching is SURF. Recent algorithms that are proposed in this field are divided into two groups. The first group have better performance than SURF but their running time have been increased. The second group reach a lower running time compared with SURF but they scarify the performance. This paper proposes a novel algorithm that has better performance than SURF with no running time overhead.