文档介绍：摘要
图像的边缘检测是图像处理和计算机视觉领域最重要的技术之一,在图像处理系统中占有重要的地位,是影响整个系统性能的关键因素。其算法的优劣直接影响整个计算机视觉系统性能的好坏。经典的基于梯度算子的边缘检测方法,如梯度算子、拉普拉斯以及拉普拉斯-高斯算子等,只适用于检测有限类型的边缘, 并且对噪声很敏感。找到与图像中目标的实际分界线相对应的真实边缘,一直是图像处理领域里的一个难题。
本文研究了基于梯度算子的边缘检测算法,针对基于梯度算子的图像边缘检测算法检测精度低、抗噪声性能差等问题提出了一种改进的 Sobel 图像边缘检测算法和一种基于模拟退火算法和梯度算子的图像边缘检测算法。其中改进的 Sobel 图像边缘检测算法在 Sobel 图像边缘检测的基础上增加了噪声检验和边缘细化步骤,克服了传统 Sobel 算子检测方向有限、受噪声影响严重等缺陷;基于模拟退火算法和梯度算子的图像边缘检测算法通过模拟退火算法获得最优梯度算子,并根据最优算子对图像进行边缘检测,克服了经典算法中使用固定算子的缺陷。
实验表明,本文提出的改进 Sobel 算子图像边缘检测算法可检测多方向的边缘,定位比较准确,抗噪声能力强并且能获得较细的边缘。基于模拟退火算法和梯度算子的图像边缘检测算法检验效果良好,稳定性高,抗噪声性能优良。
在后续的工作中,还需要深入研究两种新边缘检测算法执行效率以及所适应的边缘类型。
关键词:边缘检测梯度算子 Sobel 算子模拟退火算法
Abstract
The image edge detection is one of the most important technologies in the digital image precessing puter vision, playing an important role in image processing system. It is a critical factor affecting system's performance. The performance puter vision system is influenced directly by the accuracy of the edge detection algorithm. Most classic edge detection algorithm, such as gradient operator, Laplace operator, LOG operator, only detect edge of restricted types and they are sensitive to noise, too. It is a difficult topic in image processing research field to detect the edge correspond to the real boundary of target in image.
This paper researches the edge detection algorithm based on gradient operator and its problems, such as lower detection accuracy and noise resistance ability. An improved Sobel image edge detection algorithm and an image edge detection algorithm based Simulated Annealing and gradient operator are introduced. Sobel edge detection algorithm is influenced seriously by noise and can only detect edge of restricted types. According to these problems, the improved Sobel image edge detection algorithm add noise detection and edge thin algorithm. The image edge detection algorithm based Simulated Annealing and gradient operator use Simulated annealing to get the best operator ,which is used to detect edge. This solved the