A Difference-Based Feature Description Method of Image Target

This paper proposed a new method of feature description for target recognition and matching. Firstly, a method of calculating the difference was defined. The gray value matrix of an image was converted to a difference value matrix. Then the difference value, shape, angle and other features of a region and the combined features between regions were described. Finally, the method was applied to identify traffic signs. Experiments showed that the proposed method can represent multiple features of image such as the gray differences, the shape changes, and so on. Through theoretical and simulation analysis, even under rotation, shift or scale transformation, new features description method still can correctly recognize the target.