基于纹理平滑和GrabCut的皮影图案轮廓的智能提取
2020-12-23刘静庄梅玲石历丽高婷
刘静 庄梅玲 石历丽 高婷



摘要: 为客观有效地识别局部图案轮廓,实现可选择性目标的提取,文章以皮影图案为研究对象,针对皮影图像局部细节丰富、色彩饱和度高而背景信息干扰较大的特点,设计了皮影图案轮廓的智能提取算法。首先,采用相对总变差模型进行噪声与主结构的分离,实现图像的平滑处理;然后,设计GrabCut算法,通过分析图案轮廓的边界紧密度指标,确定最优的超像素分割数量,实现局部图案的优化分割;最后,运用Canny算子对分割后的皮影图案进行了轮廓提取。通过6幅皮影图像的轮廓提取实验结果表明,提出的方法准确完整地实现了目标图案的轮廓提取,且图案分割结果的像素准确度(PA)均大于95%。
关键词: 皮影;智能轮廓提取;相对总变差模型;超像素分割;GrabCut;Canny
中图分类号: TS941.2
文献标志码: A
文章编号: 1001-7003(2020)11-0020-08
引用页码: 111104
Abstract: In order to identify the local pattern contour and extract the optional target, this paper takes the shadow pattern as the research object. According to the characteristics of rich local details, high color saturation and strong interference of background information in the shadow image, an intelligent contour extraction algorithm for shadow patterns was designed. First of all, in order to achieve image smoothing, the relative total variation model was used to separate the noise from the main structure. Then, the GrabCut algorithm was designed. By analyzing the boundary compactness index of the pattern contour, the optimal super-pixel segmentation quantity was determined to realize the optimal segmentation of the local pattern. Finally, the contour of the segmented shadow pattern was extracted by Canny operator. The contour extraction experiment results of 6 shadow images showed that the method proposed in this paper could extract the contour of target pattern accurately and completely, and the pixel accuracy(PA) of the pattern segmentation was greater than 95%.
Key words: shadow play; intelligent contour extraction; relative total variation model; super-pixel segmentation; GrabCut; Canny
皮影是一門“动则成戏,静则成画”的古老艺术,蕴含着独特的美学特征和传统的文化内涵,至今已有两千多年历史[1]。皮影图案多采用仿生手法呈现,大体可以分为花草植物、鱼鸟动物及云纹水波三种图案类型。这些图案造型精美、内涵丰富,极具装饰性,在包装、服饰及家居用品[2-4]上已有大量应用。皮影艺术也极具地域性,不同的皮影图案可有效反映出所在时代和地域的文化特点及变迁。随着时代的发展,娱乐形式愈加多样化,皮影戏及皮影艺术正面临着退出历史舞台的危机。目前,对于皮影图案的研究主要集中于从艺术分析、应用及传承保护的角度进行[5-7],对于图案提取技术的研究较少涉及。……
