GPU-Based Active Contour Segmentation Using Gradient Vector Flow Zhiyu He and Falko Kuester Calit2 Center of GRAVITY University of California, Irvine zhe@uci.edu, fkuester@uci.edu Abstract. One fundamental step for image-related research is to obtain an accurate segmentation. Among the available techniques, the active contour algorithm has emerged as an efficient approach towards image segmentation. By progressively adjusting a reference curve using combi- nation of external and internal force computed from the image, feature edges can be identified. The Gradient Vector Flow (GVF) is one efficient external force calculation for the active contour and a GPU-centric imple- mentation of the algorithm is presented in this paper. Since the internal SIMD architecture of the GPU enables parallel computing, General Pur- pose GPU (GPGPU) based processing can be applied to improve the speed of the GVF active contour for large images. Results of our experi- ments show the potential of GPGPU in the area of image segmentation and the potential of the GPU as a powerful co-processor to traditional CPU computational tasks. 1 Introduction In the area of image based analysis and its related applications, segmentation is, in many cases, the starting point for further processing. The segmentation algorithm may provide the foundation for further processing, such as identify- ing features or objects that subsequently are used for the reconstruction of 3D models. Among many existing segmentation algorithms, the active contour tech- nique or snake [1] is an algorithm that uses an external force and an internal force to progressively fit a closed curve to edges, boundaries or other features of interest specified via gradient. The snake has been widely used in areas such as biomedical image analysis and further enhanced for specific problem domains. For example, Xu and Prince [2] proposed a better way of calculating the external force of the curve. This improved snake algorithm is called Gradient Vector Flow (GVF) snake and has two advantages over the original snake algorithm: (1) it is less sensitive to initialization and (2) it can move into boundary concavities. This paper introduces a hardware accelerated technique for gradient vector flow computation, utilizing the vertex and fragment units on today’s graphics pro- cessing units. Most mid-range GPUs now have a SIMD architecture and deep parallel processing capabilities on the vertex and fragment units [3], which can be used as a very efficient co-processor that can take over some of the computation G. Bebis et al. (Eds.): ISVC 2006, LNCS 4291, pp. 191–201, 2006. c  Springer-Verlag Berlin Heidelberg 2006