基于相对总差和水平集的织物毛球分割  

Fabric Pilling Segmentation Based on Relative Total Variation and Level Set

作　　者：赵喜林[1] 

机构地区：[1]浙江工贸职业技术学院,浙江温州325003

出　　处：《浙江工贸职业技术学院学报》2017年第4期49-52,共4页Journal of Zhejiang Industry & Trade Vocational College

基　　金：温州市公益性科技计划项目(S20150024);温州市公益性科技计划项目(G20150018)

摘　　要：针对传统的基于灰度阈值分割方法分割织物图像毛球区域精度低、难以分割完整的毛球区域等问题,提出了一种基于相对总差和水平集的织物毛球分割算法。该方法首先利用相对总差的结构提取算法来去除织物表面的纹理,完成对毛球的粗分割,再利用水平集分割算法对粗分割图像进行精分割,最后利用形态学的方法对精分割结果中的候选毛球区域进行筛选和融合,得到最终的毛球区域。结果表明,该方法与传统方法相比能够提高对织物图像毛球区域分割的精度,具有一定的实用性。The methods based on transform domain for texture removing have a certain regularity requirements and the methods based on gray value threshold can hardly segment the fabric pilling regions. To solve these two problems, this paper proposes a fabric pilling segmentation algorithm based on relative total variation and level set. This method first uses the structure extraction algorithm based on relative total variation to remove the textures of the fabric and then completes the coarse segmentation for pilling. Secondly,the method use the level set segmentation algorithm to segment the coarse segmentation result and gains the fine segmentation result.Finally, the method uses morphology method to filter and fusion the candidate pilling regions in the fine segmentation and then gains the ultimate pilling regions. The experimental results show that the method can accurately segment the fabric pilling areas with a certain practicality.

关 键 词：毛球分割 相对总变差模型 水平集 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]