基于灰色关联分析的灰度图像边缘检测研究

发布时间：2018-11-16 09:20

【摘要】：本文基于灰色关联分析可以进行边缘检测这一理论依据,研究了基于灰色关联分析的传统边缘检测算法原理,改进了传统算法中存在的抗噪性能差、阈值设置主观性强等缺陷。文中首先介绍了边缘检测的相关背景知识以及基于灰色关联分析的传统边缘检测理论基础及研究进展。其次,针对传统算法存在的缺陷进行改进。加入中值滤波器对待检测图像进行平滑滤波,增强算法的抗噪性能;结合人眼视觉特性提出了一种自适应的计算阈值的微分方程,该方程由图像待检测像素点周围3×3邻域的平均灰度值组成,克服了传统算法阈值设定主观性强的缺陷;对改进算法进行仿真分析,通过处理边缘点的八邻域区域点,改善了改进阈值提取出的边缘伪边缘较多现象;实验仿真证明,改进算法对含有较高浓度的椒盐噪声有很好的抑制效果,自适应阈值提取出的边缘较传统算法定位误差更小。最后,主要通过实验仿真,分别从抗噪性能、定位误差、线性连接程度、边缘连续性等方面,将改进算法与经典微分算子的检测性能进行对比。实验数据证明,改进算法相比经典算法而言检测出的边缘图像更为完整,边缘连续、较细,定位精度较高,对于含有较高浓度的椒盐噪声图像,也取得了经典算法不可比拟的优势。探讨了改进算法对不同类型图像的适用性以及算法时间在不同灰度级图像中的性能,结果证明本文算法适用性强。
[Abstract]:Based on the theory that the gray correlation analysis can detect the edge, this paper studies the principle of the traditional edge detection algorithm based on the grey correlation analysis, and improves the shortcomings of the traditional algorithm, such as poor anti-noise performance and strong subjectivity of threshold setting. This paper first introduces the background knowledge of edge detection, the theoretical basis and research progress of traditional edge detection based on grey correlation analysis. Secondly, the defects of the traditional algorithm are improved. The median filter is added to the detection image for smoothing filtering to enhance the anti-noise performance of the algorithm. In this paper, an adaptive differential equation for calculating threshold is proposed based on human visual characteristics. The equation is composed of the average gray value of 3 脳 3 neighborhood around the pixels to be detected in the image, which overcomes the subjective disadvantage of the traditional threshold setting algorithm. The improved algorithm is simulated and analyzed. By dealing with the eight neighborhood region points of the edge points, the phenomenon of more pseudo-edge of the edge extracted by the improved threshold is improved. The experimental results show that the improved algorithm can suppress the salt and pepper noise with high concentration, and the edge extracted by the adaptive threshold is less than the traditional algorithm. Finally, the improved algorithm is compared with the classical differential operator in the aspects of anti-noise performance, location error, linear connection degree, edge continuity and so on. The experimental data show that the edge image detected by the improved algorithm is more complete, the edge is continuous, the edge is finer, and the localization accuracy is higher than the classical algorithm, for the image with high concentration of salt and pepper noise, The advantages of the classical algorithm are also obtained. The applicability of the improved algorithm to different types of images and the performance of the algorithm time in different gray-scale images are discussed. The results show that the proposed algorithm has strong applicability.
【学位授予单位】：西安科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41

【参考文献】