Study Of Segmentation Algorithm Of Texture Image Based On Wavelet Transform

Image segmentation is an important task in many digital image processing applications.Texture provides important information in many images,especially in some natural images,and sometimes it is the major discrimination factor to image segmentation.Wavelet transform is a powerful tool in the signal and image processing,and its fundamental theory has been formed. Wavelet transform has many properties which other analysis techniques(for example,Fourier analysis,Gabor transform,etc)do not have, The coefficients in each wavelet channel of a multiresolution decomposition of an image reflect some characteristics of the image..Especially,for texture image the texture energy in each channel represents the thickness and orientation of the texture.This dissertation introduce wavelet analysis to image segmentation.Based on analysising deeply characteristics of wavelet transform,The author has made qualitative texture analysis using wavelet frames in this paper. For the object segmentation in textured background, a feature image segmentation algorithm is proposed according to the difference of texture energy distribution between textured background and objects in wavelet channels.Then,the algorithm is modified to apply to texture segmentation and object segmentation in multi-textured background.It is showed that the algorithm gives a good feature representation not only between different textures but also texture and objects.And the correct segmentation result are obtained in our experiments.The dissertation consists of five chapters. The chapter one is an introduction, which includes: the statement of problem and the significance of study, a survey about the history of image segmentation, and finally, a presentation of the main research works in this dissertation.In the chapter two, from the point of view of signal and image processing, some basic theories about wavelet transform are introduced. This chapter set forth the definition and the character of wavelet transform, binary wavelet and wavelet frame. The kernel of wavelet transform, i.e. multi-resolution analysis, the decomposition reconstruction method of sign and image multi-resolution, wavelet feature in the image process are all discussed.The chapter three is about texture image segmentation and texture analysis. The definition of texture and normal texture segmentation method are introduced. By comparing different texture image segmentation method, this chapter presents a method of texture image segmentation using discrete wavelet transform.The chapter four is about segmentation algorithm-feature image method. This chapter presents the feature image segmentation algorithm. The algorithm uses the wavelet channel's texture energy value that produced by discrete two dimension wavelet decomposition as the description feature. Furthermore this chapter advances the algorithm and applies it to texture image segmentation and object segmentation in multi-texture background. Finally complete the final segmentation by using domain merging algorithm and introducing dominative image. The experimental result shows that it is a very effective image segmentation algorithm. The chapter five is the conclusion, which gives the main research result.