边缘检测算法及其在面瘫识别系统中的应用

发布时间：2018-04-08 10:41

本文选题：图像处理　切入点：边缘检测　出处：《长春工业大学》2017年硕士论文

【摘要】：边缘检测作为数字图像处理中一种常用的处理方法,是很多图像处理技术的基础。图像的边缘包含了大量的图像特征信息,可用于特征描述、图像增强、图像分割、模式识别等图像分析和处理中。因此图像边缘检测算法一直是图像处理和计算机视觉领域的研究热点,期望寻找一种检测精度和抗噪能力都令人满意的算法。面瘫是一种人脸面部神经运动功能受损的疾病,它给人的身心都带来极大的伤害。但现在医学对面瘫的判定大都基于主观判断,没有一个统一的评价标准。利用边缘检测算法建立面瘫识别系统,准确判定面瘫程度,对于提早发现疑似病例、指导临床用药、评估治疗效果以及学术领域交流等具有重要的意义。本文利用边缘检测算法对面瘫图像进行了处理,期望建立一个较为标准的面瘫识别系统,从而实现对面瘫图像的识别与分级。本文主要完成改进边缘检测算法和建立面瘫识别系统两方面的工作。本文首先对一些传统的边缘检测算法进行了分析与研究,并介绍了三种新兴的边缘检测方法,在对这些方法归纳研究的基础上,提出了一种改进的边缘检测算法。通过增加模板方向来改进Sobel算法,再利用遗传算法改进Canny算法的阈值选取过程,然后采用小波变换的方法对两种改进的算法进行图像融合处理,结合两种算法的优点,得到改进的边缘检测算法。然后,针对面瘫图像特征的位置信息,设计了一种面部图像采集系统,获取含有区域分块的人脸图像。在建立面瘫识别系统时,利用改进的边缘检测算法对面瘫图像进行处理以提取图像边缘特征信息。根据面瘫患者面部不对称的性质,计算特征信息矩阵之间的差值,提出一种基于对称轴的面瘫分级方法。再将其与基于距离的面瘫评分方法相结合,完成对图像最终的等级划分,建立面瘫识别系统,实现对面瘫图像的识别与分级。
[Abstract]:Edge detection, as a common processing method in digital image processing, is the basis of many image processing techniques.The edge of the image contains a large amount of image feature information, which can be used in image analysis and processing such as feature description, image enhancement, image segmentation, pattern recognition and so on.Therefore, image edge detection algorithm has been a hot topic in the field of image processing and computer vision. It is expected to find an algorithm with satisfactory detection accuracy and anti-noise ability.Facial paralysis is a kind of facial nerve function damage disease, it brings great harm to the body and mind.But now the medical judgment of facial paralysis is mostly based on subjective judgment, without a unified evaluation standard.Using edge detection algorithm to establish facial paralysis recognition system and accurately determine the degree of facial paralysis is of great significance for the early detection of suspected cases, the guidance of clinical medication, the evaluation of therapeutic effect and the exchange of academic fields.In this paper, the edge detection algorithm is used to deal with the facial paralysis image, and it is expected to establish a standard facial paralysis recognition system, so as to realize the recognition and classification of the facial paralysis image.In this paper, the improvement of edge detection algorithm and the establishment of facial paralysis recognition system are mainly completed.In this paper, some traditional edge detection algorithms are analyzed and studied, and three new edge detection methods are introduced. Based on the research of these methods, an improved edge detection algorithm is proposed.The Sobel algorithm is improved by adding the template direction, and the threshold selection process of the Canny algorithm is improved by genetic algorithm. Then the image fusion of the two improved algorithms is carried out by wavelet transform, which combines the advantages of the two algorithms.An improved edge detection algorithm is presented.Then, according to the position information of facial paralysis image feature, a facial image acquisition system is designed to obtain the face image with regional block.In order to extract the edge feature information of facial paralysis image, the improved edge detection algorithm is used to process the facial paralysis image.According to the character of facial asymmetry in facial paralysis patients, the difference between characteristic information matrix is calculated, and a method of facial paralysis classification based on symmetry axis is proposed.Then combining it with the distance based facial paralysis scoring method, the final classification of the image is completed, and the recognition system of facial paralysis is established to realize the recognition and classification of the facial paralysis image.
【学位授予单位】：长春工业大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】