| Image analysis in nature environment is widely applied in many fields. In recent decades, many specialists and scholars have employed themselves studying target detection and recognition and lots of new techniques have appeared. One of them is fractal which is used in image processing mainly based on two points: one is that nature environment accords with fractal Brown motion model in a large range; the other is that fractal dimension are analyzed as a very important image parameter.In this paper, fractal technique is studied and used in image edge detection and image segmentation. The Shortcomings and advantage of it are summarized and an improved method is proposed as follows. A new detection arithmetic based on fractal error is presented which is superior to traditional brim detection method. First, image data is transformed from gray space to fractal characteristic space which is also called error space. Second, the difference of the dissimilar fractal models which are indications of target and background is used to edge detection. The value of fractal error is higher and it's corresponding pixel is more consistent with FBM model. As a result, the pixel is in more possibility belonged to brim parts. The simulation results indicate the algorithm has the excellent performance in anti-noise and little target or multi-target detection.Image segmentation method is proposed which is based on region fractal and watershed technique. Compared with other segmentation methods, the new one has good precision in orientation and outstanding effect, which can restrain and eliminate the influence of natural background especially in complicated natural background and single target character. It is also an novel idea for resolving the problems of artificial target detection and region fractal.On the whole, the methods in this paper are superior to conventional arithmetics, which is also an novel idea for resolving the problems of artificial target detection and recognition in nature environments. |
PDF Full Text Request