摘要
轮廓提取在许多智能视觉系统中(特别是在模式识别中),被认为是非常重要的过程。图像轮廓提取首先要进行边缘检测,然后提取阶跃结构的边缘,这2个步骤一般情况下是分开进行的。介绍的尺度独立算法是建立在多尺度分析与MAS小波变换理论的基础上,并结合数学上描述函数奇异性的Lipschitz指数知识,它不仅能根据梯度方向上的局部最大振幅有效进行图像边缘检测,还能根据MAS小波变换后的图像梯度向量振幅与变换尺度的无关性,提取出作为图像轮廓的阶跃结构边缘,并能有效消除噪声。
Finding contour in an image is considered to be an important process in many artificial vision systems, especially in pattern recognition. Edge detection must be firstly processed in image contour extraction, then the step-structure edge that is regarded as image contour should be extracted. Generally, two steps are detached. Scale-Independent algorithm in this paper is based on multiresolution analysis and MAS wavelet transform theory, and combined with Lipschitz exponent knowledge which is a remarkable mathematical tool to analyze the singularities including the edges. It can not only detect image edge according to the local maximum modulus along the gradient direction, but also easily to extract step-structure edge of image and remove the noise according to the modulus of the MAS wavelet transform that is independent of the scale of the transform.
出处
《重庆邮电学院学报（自然科学版）》
2004年第2期44-48,共5页
Journal of Chongqing University of Posts and Telecommunications(Natural Sciences Edition)
基金
重庆市科委应用基础研究基金项目(6799)
关键词
边缘检测
轮廓提取
多尺度分析
MAS小波变换
edge detection
contour extract
multiresolution analysis
MAS wavelet transform
