Aiming at how to establish the ideal standard for the edge detection, a new image edge detection method was proposed based on a combination of least squares support vector machine (LSSVM) and cellular automata. Polynomial and Gaussian kernel function was deployed to construct a new kind of kernel function. LSSVM selected the new kernel function and fitted the image intensity surface for the neighborhood of every pixel. The gradient operators which were deduced from the above LSSVM convoluted with the image gray values to get the image gradient values. Gradient values were evolved out by cellular automata with the designed local rules in order to achieve the best edge detection performance. Simulation results showed that edges were a pixel width and edge positioning was accurate. As illustrated that the proposed algorithm was feasible. Furthermore, the proposed algorithm was higher than the Sobel and the Canny algorithm in detection performance.
