Abstract: In allusion to the craftsmanship and structure of pattern primitive of Xinjiang carpet, a pattern edge extraction method which combines color clustering and edge detection is put forward. According to the need to determine a processing area, Laplace filter of the [-8] center is used to enhance the image. K-means color clustering is adopted to regionally cluster the enhanced targets. On the basis of color, target area is segmented. Canny operator is applied to detecting the close edge of pattern primitive, and it is indicated with compressed Freeman chain. The experimental results show that the proposed method for Xinjiang carpet with diversified pattern has preferable effect and establishes a favorable foundation for digital processing of national symbols.

Key words: image enhancement, K-means clustering, edge detection, compressed Freeman chain

摘要： 针对新疆民族地毯的制作工艺及图案基元的构造，提出一种结合颜色聚类与边缘检测的图案边缘提取方法。根据需要确定待处理区域，使用中心为[-8]的拉普拉斯滤波器对其增强；采用K-means颜色聚类对增强后的目标区域进行聚类，根据颜色分割出目标基元的大致轮廓；运用Canny算子进行边缘检测提取出图案基元闭合边缘并采用Freeman链码的压缩形式进行表示。实验结果表明，该方法对新疆民族地毯多样化的图案基元的提取具有较好的效果且为民族元素数字化处理奠定了良好的基础。

关键词: 图像增强, K-means聚类, 边缘检测, Freeman链码压缩