SURF (Speeded-Up Robust Features) is a scale-and rotation-invariant algorithm, which has a better repeatability, distinctiveness, robustness, and a faster computing and comparing speed. In this paper, we propose an improved SURF algorithm based on ACO (Ant Colony Optimization). First of all, the algorithm uses SURF to find all interest points. Secondly, each pixel of the original image is seen as an ant, imitates the process of ants search food to get the image edge. Finally, selects the

more »

... t points from the area around the image edge. The experimental results show that the interest points extracted by the improved SURF algorithm based on ACO are more robust, and the number of them is effectively reduced. In this way, we can reduce the amount of calculation for the subsequent image registration. Introduction Image registration is a hot issue in the field of remote sensing data analysis, computer vision, and image processing, and so on. It is a process of two or more images matching or overlaying, the images are obtained in different time, different sensors (imaging equipment) or different conditions (weather, illumination, camera position and angle, etc.). At present, it is widely used, including 3D reconstruction, image retrieval, object recognition, and so on. Generally, image registration can be classified into image registration based on gray scale and image registration based on features. Image registration based on gray scale usually directly uses gray information of the whole image, and thinks that corresponding points and surrounding areas of corresponding points between images are with the same or similar gray value, so it makes gray similarity the basis of the definition of similarity measure, then designs corresponding search strategy to find the optimal image geometry transform parameter which can make similarity measure function reach a maximum. For the reason it needs to calculate the gray value of the points in the area around the matching points, its computation amount is large, and also has a slower speed. While image registration based on features extracts features remain unchanged in the image, such as edge point, the center of closed region, and so on, as the reference information of image registration. Using this kind of method, the main advantage is that it extracts remarkable features of the image, greatly compresses the image information, and has good robustness, small amount of calculation, fast speed. Image registration based on features has already become the mainstream of the image registration research direction [1]. SURF (Speeded-Up Robust Features) [2], proposed in 2006 by Herbert Bay, is one of the feature-based image registration method, including two parts, interest point detection and interest point description. This algorithm is proposed on the basis of SIFT (Scale-Invariant Feature Transform) algorithm [3] . The same as SIFT, the features extracted by SURF also have scale-and rotation-invariant performance, and with respect to illumination change, affine and perspective transformation, it has partial invariance. In repeatability, distinctiveness, and robustness these three aspects, SURF all exceeds or approaches the previously proposed such kind of methods, and there is obvious advantage in the speed of computing [4] . In this paper, we use SURF to find interest points, and then use ACO (Ant Colony Optimization) [5] to detect image edges [6], select the interest points around the area of the image edge as the final interest points. The next section introduces interest point extracting based on SURF. The third section