【作者】 乐宋进；

【导师】 胡泳芬； 武和雷；

【作者基本信息】 南昌大学 ， 电机与电器， 2005， 硕士

【摘要】 简单地说,图像分割就是把图像中的物体与背景或物体与物体分割开。它是图像识别与理解中的关键步骤之一,分割质量的好坏将直接影响识别和理解的结果。近年来,在图像处理和计算机视觉领域中,符合人眼视觉模型的多分辨率技术日益受到人们的重视。在图像分割的研究中,基于人类视觉感知特性的图像分割方法也是一个新的重要研究方向。具有“数学显微镜”美誉的小波分析,被认为是目前能够模拟人眼感知方向性和多尺度特性的最佳数学工具。 本文在对图像分割技术进行综合研究的基础上,利用小波变换多分辨率分析的方法和小波变换的特殊性质,改进以前依靠窗口大小改变的基于小波包变换,本文把小波包变换和能量检测算法相结合起来图像处理方法,通过能量检测算法来改变以前小波包依赖窗口大小局限性。小波包变换是小波变换的推广,它继承了小波变换适合于处理非平稳信号的特点,又克服了小波变换只能对信号的低频成分作进一步细分的缺点。 本文的主要工作如下: 讨论了图像信号和噪声在小波系数上的不同分布,为在小波变换域中利用阈值设置来区分二者提供了理论依据。在此基础上提出了一种自适应图像去噪方法,结合图像的自身特征对噪声和信号加以区分,通过调整因子的设置增强图像边缘,对提出的去噪方法进行了实验仿真分析,取得了较好的增强效果,在降噪的同时保持了大部分细节及边缘信息。 分析小波变换和小波包变换的优缺点,改进了一种基于小波包变换的图像分割方法,并给出了仿真实例,实验结果证明了该方法的有效性。更多还原

【Abstract】 Image segmentation, which is to segment objects and background in image from one another, is an essential and important topic in image recognition and image understanding. In recent yeas, multiresolution is being increasingly used in image processing and computer vision and it is also a new effective way to solve the problem of image segmentation. As "mathematical microscope", wavelet analysis become an new tool to simulate visual perception of multi-scale for its feature of multiresolution from coarse to fine.This paper improved on image of wavelet pack transform which has been depend change of window , it is based on colligation image segmentation, it is utilized method of wavelet transform and special representation of the wavelet analysis. We propose a image segmentation which has combined wavelet pack and arithmetic of energy examine, with energy examine, it has changed localization with energy examine, It is a modified wavelet transform, which is suitable to process non-stationary signal but can performs further decomposition only in low frequency. With the wavelet packet transform, we are able to zoom into any desiredfrequency channels for further decomposition.The main research results in this paper were as follows:The different distribution of image signal and noise versus wavelet coefficient werediscussed. This research provided the theory basis for separating signal from noise by threshold. An self-adaptive local threshold scheme was proposed. In combination with natural characteristics of image, the signal could be distinguished in wavelet transform, the edge of image could be enhanced with adjusting factor. Simulation results were provided in this paper. Satisfying results were obtained in noise smoothing, with image edge and other main features being retained. The advantages and disadvantages of wavelet transform and wavelet packet transform were analyzed, an image segmentation method was improved with the basis of a wavelet transform packet. Simulation results were provided in this paper. The research results proved the efficiency of the method.更多还原