Image Segmentation Based On Fractal Theory

In the dissertation, author reviews the developed history and presents situation of fractal theory, and introduces Mandelbrot' s theory of fractal and primary conception of image segmentation, makes a comparison among some fractal models and some methods to acquire the fractal dimension of a image in complexity and applicability, advances a kind of new differential box counting (DBC) method, in which scale of box is variable adaptively. In the dissertation, author has a conception of frequency domain fractal and discusses it primarily.Author calculates the fractal dimension of natural and man-made target in infrared image by a few kinds of methods, and makes use of some related ways, such as Edge Preserving Noise Smoothing Quadrant (EPNSQ) filtering approach and neural network to segment images, and uses VC++6.0 and the way of combination programming between VC++ and Matlab to accomplish the segmentation.In the dissertation, author discusses the relations of fractal with chaos and of fractal with wavelet, and introduces briefly the application of fractal in digital image watermarking techniques. On the end of the dissertation, author points the developing direction of the project, looks forward to the future of fractal theory and gives significative view.