摘 要
毕业论文题目： 彩色图像边缘检测算法及应用研究
随着彩色图像的普及和应用，彩色图像处理已引起人们的关注。目前国内外已经逐步对彩色图像边缘检测算法进行深入研究，针对不同色彩空间、不同图像处理框架、张量理论提出了一些应用算法，传统的灰度图像边缘检测算法没有充分考虑图像的色彩信息和人眼的视觉感知特性，通用性不够而没有被广泛应用，算法具有一定的局限性，因此研究彩色图像边缘检测算法及应用具有重要的意义，同时也是完成《基于CSLIP算法的路况视频能见度检测应用研究》(编号：BK2010366)、《基于人眼视觉重建技术的交通能见度检测方法研究》(编号：BE2011511)等省级项目的理论依据。
论文研究了彩色图像边缘检测算法及其应用，优化和改进了传统的灰度域边缘特征提取算法。论文主要工作包括：介绍了色彩空间和人眼视觉理论；提出了利用色彩张量和光照准不变微分式的图像特征提取算法，阐述了不同色彩空间彩色图像边缘特征的光照准不变微分式，结合色彩张量算子提取彩色图像的阴影-遮挡准不变边缘、镜面反射准不变边缘和阴影-遮挡-镜面反射准不变边缘，最大程度地降低彩色图像边缘检测的误检率和漏检率，有效提高边缘检测器的性能；提出了基于对比敏感度函数CSF的自适应局部色差可视阈值的边缘检测算法，构建背景亮度掩膜与对比灵敏度的色差可视阈值影响函数，结合局部背景亮度及空间频率对人眼视觉的影响，解决了目前基于人眼空间频率特性进行彩色图像边缘检测算法的难题；提出了修正后的单一因子LIP算法，改进已有的参数化LIP框架，设计单一因子构建LIP与传统算术运算之间的关系，将此单一因子LIP图像处理框架应用于视频图像的纹理特征描述，检测结果更加符合人眼的视觉感知，提高了边缘检测器的抗噪性能；给出了基于四元数理论的彩色图像边缘检测算子，利用四元数和矩阵理论，构造四元数边缘检测算子，能更好的保留彩色图像的轮廓特征，更具有实际意义。
经不同场景、光照等彩色图像序列的测试，论文提出的彩色图像边缘检测算法在检测准确率、检测精度、抗噪性能上优于其他已有算法。论文研究成果已在沪宁、崇启等高速路网视频监控综合系统及能见度检测系统中试用。
论文主要创新或特点在于：
? 提出了利用色彩张量和光照准不变微分式的图像特征提取算法，降低了彩色
I
摘 要
图像边缘检测的误检率和漏检率，有效提高边缘检测器的性能；
? 提出了基于对比敏感度效应CSF的自适应局部色差可视阈值的判断方法，解
决了目前基于人眼空间频率特性进行彩色图像边缘检测算法的难题； ? 提出修正后的单一因子LIP算法，将此单一因子LIP图像处理框架应用于视
频图像的纹理特征描述，结果更加符合人眼的视觉感知；
? 给出了基于四元数理论的彩色图像边缘检测算子，能更好的保留彩色图像的
轮廓特征，更具有实际意义。
关键词：色彩空间；人眼视觉感知；色彩张量；CSLIP图像处理框架；四元数
理论；彩色图像边缘检测
II
ABSTRACT
THESIS：The Study of Color Image Edge Detection Algorithms and
Its Applications
Color image processing is playing a more and more important role with the popularization and application of color images. At present, many countries at home and abroad have gradually begun to study on the subject--color image edge detection algorithm. According to different color spaces, color image processing frameworks and color tensor theory there are some applications, but mostly they are not widely used due to some disadvantages such as bad real-time property or not enough commonality. Therefore, the study of color image edge detection algorithms and application is imperative and is also the theoretical basis of these Provincial projects:
The paper presents color image edge detection algorithms which improves the traditional gray edge detection methods. Paper main jobs include: the color space and human visual theory, put forward the color image feature extraction algorithm based on color tensor and light quasi-invariant differentials in different color spaces to decrease the color image edge detection rate and effectively improve the performance of edge detection; put forward a color image edge detection algorithm based on local adaptive perceptual color difference, combine the influence of local changes in luminance and spatial frequency to human visual system to fill in the blank of color image edge detection algorithm based on human visual perception; put forward the single factor LIP algorithm to improve the existing parameters of LIP framework, design a factor to construct the relationship of LIP and traditional arithmetic, apply this algorithm in video image texture feature descriptions, the results are more consistent with human visual perception; present the color image edge detection operator based on quaternion theory, use quaternion and matrix theory to construct quaternion edge detection operator to acquire color image contour features and has more practical significance
According to the test in different scenes and light conditions, the color image edge detection algorithms proposed in this paper are superior to other existing algorithms in detection accuracy and noise resistance. The paper’s research results have succeeded applying in include Shanghai and Nanjing’s high speed road network video surveillance system and visibility detection system.
The paper’s main innovation spots:
? Put forward the color image feature extraction algorithm based on color tensor and light
III
通信工程论文 彩色图像边缘检测算法及应用研究
