Method Of Color Image Segmentation Based On Wavelet Theory

Image segmentation is thoroughly active research domain in the field of computer image recognition and comprehension. The realizable method of computer image comprehension is based on image segmentation. In the recent years, the specialists and scholars had attached importance to the color image segmentation which is a more and more important method of image segmentation.The main contribution and valuable results of this dissertation can be listed as following:At the beginning, through analyzing the process of color image segmentation and comparing the results of 'color image segmentation using fuzzy C-means and eigenspace projections' and 'color image segmentation using histogram multithresholding and fusion' ,the advantages and drawbacks of these two methods are found. The arithmetic of them are too complex and the time used to account are too long even though they both can get good segmentation effect.And then, to deal with the deficiencies of the above arithmetic, a novel arithmetic of the color image segmentation is proposed, which is based on the complete and systemic research of wavelet theories. This arithmetic is called color image segmentation using histogram multithresolding based on planar wavelet, it combined with the characteristics of wavelet transform to overcome the deficiencies of the above arithmetic. It not only decreases the complexity of arithmetic and saves the accounting time, but also owns powerful noise resistance due to considering the information of color and space to image simultaneity.In the end, the evaluation frame which is based on the research of evaluation method of the color image segmentation is proposed. To validate the superiority and feasibility of all arithmetic, two groups of test models are designed. The results of simulation to these test models and other images by using MATLAB language testify their applicable occasion, color image segmentation using fuzzy C-means and eigenspace projections is adapt to apply to the occasion of the noise is small,the degree of illegibility is small and the size of objects is large enough, color image segmentation using histogram multithresholding and fusion isn't sensitive to noise and the size of objects,so it suits to segment the image with noise and the size of objects is small,when the size of objects is lagre enough, color image segmentation using histogram multithresolding based on planar wavelet is the smallest suffered from noise and illegibility,so, it is adaptive to segment images with noise and illegibility, that is, it's robust to the condition of sampling.