基于Gabor滤波器的超声图像的边缘增强和边缘检测

发布时间：2018-01-22 20:36

本文关键词： Gabor滤波器超声图像边缘检测边缘增强　出处：《内蒙古大学》2017年硕士论文　论文类型：学位论文

【摘要】：本文的主要研究内容是对B超图像和声呐图像的边缘进行增强与检测。在医学领域,医生依靠对超声图像的观察来判断患者病情;在海洋领域,声呐员通过对声呐图像的分析来探究海洋。但是一般B超图像和声呐图像的边缘分辨率低、品质恶劣,而图像的边缘又是图像中的重要信息,如果不进行处理,会严重影响研究者的判断。因此,本文结合B超图像和声呐图像的特点并根据现有的边缘增强和边缘检测的理论,进一步探究基于Gabor滤波器的超声图像的边缘增强和边缘检测的方法。为解决直方图均衡化中增强的对象是图像的全局而不是图像中的目标本文使用将直方图动态分割为多个子直方图的方法,并且提出了在直方图曲线的波峰和波谷1/2处划分的方法;为解决直方图均衡化在增强过程中图像中高频灰度数量增加、低频灰度数量压缩会导致图像出现不自然的过度增强现像,本文又将分割完的子直方图进行了自适应三角形的对比度增强算法(Adaptive Trilateral Contrast Enhancement,ATCE)。这样,既使直方图均衡化的方法具有目标性,又减轻了因灰度级分配不均匀而导致的过度增强现象。在实现图像边缘锐化的方法中,本文使用了修正的Gabor滤波器(modified Gabor filter,MGF)与高频补偿滤波器相结合的方法。Gabor滤波器的方向和频域表达类似于人眼系统,并且符合海森堡测不准原则所判定的有效时间与有效频率带宽两者乘积的下限。但Gabor滤波器获得的频率和方向有限得到的边缘信息模糊且残缺不全,本文通过使用具有双波峰的MGF使检测出的边缘更清晰更完整,再利用高频补偿滤波器对图像的边缘进一步锐化和去噪。在实现图像边缘检测的方法中,本文提出了基于矩形核Gabor滤波器(Rectangular Nuclear Gabor Filter,RNGF)的边缘检测方法,该方法兼具矩形核对方向判断准确的优势和Gabor滤波器自身的优势,在一定程度上解决了使用矩形核高斯函数的边缘检测方法易出现噪声的问题。
[Abstract]:The main content of this paper is the enhancement and detection of ultrasound images and sonar image edge. In the field of medicine, doctors rely on observation of ultrasound images to determine the patient's condition; in the field of marine sonar who, through the analysis of the sonar image to explore the ocean. But the general B images and sonar image resolution is low. Poor quality, and the edge of the image is an important information in the image, if it is not treated, it will seriously affect the researcher's judgment. Therefore, this paper combined with the characteristics of ultrasound images and the sonar image and according to the existing edge enhancement and edge detection theory, to further explore the method of ultrasound image edge enhancement and Gabor filter based on edge detection. In order to solve the problem of object enhanced histogram equalization is the image of the whole and not the object in the image is used in this paper will be divided into a plurality of sub dynamic histogram The histogram method, and put forward a method in the histogram curve peaks and trough 1/2 division; in order to solve the problem of histogram equalization in high frequency enhanced gray image in the process of quantity increase, low frequency gray compression will lead to excessive number of image enhancement is not a natural phenomenon, this paper will end by sub histogram contrast adaptive enhancement algorithm (Adaptive Trilateral Contrast Enhancement triangle, ATCE). In this way, even if the method of histogram equalization is targeted, but also reduce the gray level due to uneven distribution caused by excessive enhancement. In method of image edge sharpening, we use a modified Gabor filter (modified Gabor filter. MGF) direction and frequency domain method.Gabor filter combined with high frequency compensation filter expression similar to the human system, and in line with the original Heisenberg uncertainty The lower limit of effective time is determined and the effective frequency bandwidth of both product. But the Gabor filter to obtain the frequency and direction of the finite fuzzy edge information and incomplete, this paper through the use of double peak MGF has the edge detected is clearer and more complete, with high frequency compensation filter for image edge sharpening and go further noise. The implementation method of image edge detection, this paper presents rectangular filter based on nuclear Gabor (Rectangular Nuclear Gabor Filter, RNGF) edge detection method, this method has advantages of accurate direction of rectangular check and Gabor filter their own advantages, to a certain extent to solve the edge detection method using rectangular kernel Gauss function prone to noise problems.

【学位授予单位】：内蒙古大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】