基于语义分割的简洁线条肖像画生成方法  

Concise line portrait generation method based on semantic segmentation

作　　者：吴涛 董肖莉 孟伟[1,2] 徐健 覃鸿[3,4] 李卫军[3,4] WU Tao;DONG Xiaoli;MENG Wei;XU Jian;QIN Hong;LI Weijun(School of Information Science and Technology,Beijing Forestry University,Beijing 100083,China;Engineering Research Center for Forestry-oriented Intelligent Information Processing of National Forestry and Grassland Administration,Beijing 100083,China;Institute of Semiconductors,Chinese Academy of Sciences,Beijing 100083,China;School of Microelectronics,University of Chinese Academy of Sciences,Beijing 100045,China)

机构地区：[1]北京林业大学信息学院,北京100083 [2]国家林业草原林业智能信息处理工程技术研究中心,北京100083 [3]中国科学院半导体研究所,北京100083 [4]中国科学院大学微电子学院,北京100045

出　　处：《智能系统学报》2021年第1期134-141,共8页CAAI Transactions on Intelligent Systems

摘　　要：针对目前主流的线条提取算法对于区域对比度不明显的边缘的检测能力较弱,且对于所有区域采用无差别、统一化的处理策略,所生成的线条画往往较复杂,非常不利于机器人机械臂绘图的问题,本文提出了一种基于语义分割的简洁线条肖像画生成方法(concise line portrait generation based on semantic segmentation,CLPG-SS)。首先,对人脸图像进行语义分割,将人脸划分为不同的区域,基于不同区域提取边缘轮廓与五官细节线条,进行边缘切向流优化,从而加强方向信息;在此基础上,利用线条图来生成调和图像,并利用优化后的边缘切向流、人脸语义分割结果以及调和图像,针对不同的分割区域调整线条提取方法的参数,实现对细节无关区域的线条过滤和细节重点区域的线条加强,生成简洁线条肖像画。实验结果表明:本文提出的CLPG-SS方法能够有效提取人脸主轮廓线条,并针对不同区域实现了对细节线条的针对性调节,提高了机器人机械臂的绘制效率。Currently,mainstream line extraction algorithms have weak detection capabilities for edges with inconspicuous regional contrast,and they use an undifferentiated and unified processing strategy for all regions.The generated line drawings are often complex,which is very unfavorable for robot manipulator drawing.Given this situation,this paper proposes a concise line portrait generation based on the semantic segmentation(CLPG-SS)method.In this method,semantic segmentation is performed on the face image,and the face is divided into different regions.Edge contour and facial detail lines are extracted based on different regions,and edge tangent flow is optimized to enhance the direction information.On this basis,the line image is used to generate the harmonic image,and the optimized edge tangent flow,facial semantic segmentation results,and harmonic image are used to adjust the parameters of the line extraction method for different segmentation regions to realize the line filtering of the detail independent region and line enhancement of the detail focus region,generating a concise line portrait.The experimental results showed that the proposed CLPG-SS method could effectively extract the main contour lines of a human face,adjust the detail lines for different regions,and improve the rendering efficiency of a robot manipulator.

关 键 词：语义分割 区域轮廓 切向流场 简洁线条肖像画 人脸线条提取 基于流的高斯差分 参数调整 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]