| Texture segmentation has drawn most attention of researchers and has been one of the difficult study topics in digital image processing and pattern recognition. Texture segmentation is defined as partitioning an image into sections accurately according to the textured regions which are composed of the same texture or by recognizing the borders between different textures in the image. Different from conventional texture segmentation methods, our method combined the wave packets transform without down-sampling and geodesic active contour (GAC) model of vector valued image in partial differential equations (PDE), and proposed a texture segmentation method with good performance.Firstly, our paper reviews the basic theory of wavelet, then analyzes its property and its effect on image processing. We concluded the two-scale relationship of scale function and wavelet function, which forms the foundation of the wavelet decomposing and reconstructing algorithm.Secondly, we introduce some image segmentation methods in common use and mainly focus on the analysis of texture-based image segmentation methods. Then we summarize the merits and drawbacks of these methods.Thirdly, we depict the principle of segmentation model in partial differential equations and give emphasis on discussing and realizing the digital arithmetic of GAC.Finally, we propose a texture segmentation arithmetic based on the wave packets transform. We get the sub images via wave packets transform of original image , compose them into a vector valued image after some certain processing to the sub images, then we segment the vector-valued image using vector GAC model in PDE. The experimental results shows our method can get a similar segmentation result as segmentation of vector-valued image based on Gabor filters. This method makes good use of the merit of the fast transform wavelet, meanwhile coefficient after decomposition can be applied to object-oriented coding .Thus it can be used widely in more fields. |
PDF Full Text Request