【作者】 张海建；

【导师】 陈向东；

【作者基本信息】 西南交通大学 ， 通信与信息系统， 2008， 硕士

【摘要】 在图像处理应用中,边缘提取是一个重要的方面,许多相关的算法被相继提出。本论文主要工作是小波变换图像去噪预处理与模糊边缘提取的研究。一般来说,现实中的图像都是带有噪声的,而图像中的噪声对图像分析、图像压缩等有很大的影响,所以为了后续边缘提取的处理,有必要对图像进行去噪预处理。本文利用小波变换的多分辨率特性,提出一种改进的小波阈值去噪方法,利用小波变换后的细节图像小波系数的区域相关性和方向性,估计信号系数的方差,根据信号系数方差和噪声方差的关系来对系数进行自适应处理,在去噪的同时较好的保持了边缘,取得了很好的效果。由于人的视觉特性和数字图像本身所具有的模糊性,使得模糊技术适用于图像处理,特别是边缘检测和图像分割等方面。遗传算法采用同时处理群体中多个个体的方法,并对搜索空间中的多个解进行评估,使遗传算法有较好的全局搜索性能。因此本文将模糊理论与遗传算法引入到图像的边缘提取中,提出了一种新的基于遗传算法优化的图像模糊边缘检测方法。该算法首先根据图像确定模糊隶属度函数的窗口宽度,然后用方向模板求出图像的梯度,根据梯度图像直方图采用遗传算进行模糊阈值优化,从而得到边缘图像。对于含噪图像,处理前首先进行小波去噪预处理。结果表明,本文方法取得较好效果,并显著提高了运算速度。更多还原

【Abstract】 Image edge detection is an important aspect of image processing. Many methods are proposed by many researchers. The mainly research of this paper is the study on image denoising and edge detection based on wavelet transform and fuzzy theory.Many images have noise in actual application. Noise effects the following image operations, such as image analysis and image compression. So image denoising is necessary before the edge detectioa According to the multi-resolution analysis of wavelet transform, an adaptive wavelet threshold denoising algorithm is presented for noise removal of the image. Because different relativity of different detail subimages in region and direction, this method estimates wavelet coefficients’ variances by using different window in different detail subimages. Then the wavelet coefficients are processed by an improve threshold and reconstructed. The experiment result shows that the algorithm can maintain the details of the image while remove the noise effectively.Fuzzy technology is fit for using in image enhancement, image segment and edge detection, because the fuzzy characteristics of human vision and images. GA has better global searching ability, because it can process several individuals and estimate the result in the same time. This paper proposed a new method of image fuzzy edge detection based on genetic algorithm. First, gradient image is constructed by using direction templates convolve with image. Then, fuzzy partition to gradient image is implemented. To solve the difficulty of selecting the size of fuzzy window, a range of it is presented according to the gradient image. Genetic algorithm is used for searching the best threshold and reducing the running time. The edge detection is realized according to the segmentation of gradient image by the threshold. Experiment result shows that the proposed method has better performance on edge detection and computing speed.更多还原