Edge detection is a vital pre-processing step of many image analysis systems. In this paper, we intend to present a novel algorithm to automatically extract color image edge by integrating multi-dimensional gradient analysis and statistical analysis on local regions. Compared with previous gradient-based edge detection algorithms, our algorithm does not need to select an appropriate threshold against the gradient magnitude. With an elaborate edge detector, we exploit image pixel gradient direction, magnitude, spatial information and region property to obtain color image edge. To avoid the problem of detecting false edges, we conduct statistical analysis on certain local regions, which have high edge density, to further optimize our edge detection result. Experimental results demonstrate the performance of our algorithm on different color images.