摘要
提取人体着装图像的服饰区域时,易受光照、阴影遮盖与人体姿态、肤色等问题的影响,提出一种融合自适应局部特征与改进模糊C均值(fuzzy C-means,FCM)的服饰图像分割算法。首先,通过改进区域生长法消除阴影实现前景提取;其次,采用MDSMGR-WT超像素分割获取图像局部特征,将自适应局部信息融入双加权FCM目标方程中,实现二次精细化分割;最后,经肤色检测提取目标服饰区域。实验结果表明,该方法的准确率可达78.93%,召回率90.12%,查准率89.93%。该方法能够减少内部区域噪声,提高服饰图像的分割精度。
Aiming at the problems that the target clothing area of the dressed human body images is easily affected by the environment illumination,shadow covering,human posture and skin color,a dothing image segmentation method with adaptive local features and improved fuzzy C-means(FCM)was proposed.Firstly,improved region growing method was used to eliminate shadows for foreground extraction.Then local features of images were obtained by MDSMGR-WT superpixel segmentation and adaptive local information was incorporated into the double weighted FCM target equation to achieve secondary refinement segmentation.Finally,the target clothing area was extracted by skin color detection.The experimental results show that the accuracy of this method can reach 78.93%,recall is 90.12%,and precision is 89.93%.It can reduce the internal area noise and improve the segmentation accuracy of clothing image.
作者
李立瑶
顾梅花
杨娜
LI Liyao;GU Meihua;YANG Na(School of Electronics and Information, Xi’an Polytechnic University, Xi’an 710048, China)
出处
《纺织高校基础科学学报》
CAS
2021年第1期47-54,共8页
Basic Sciences Journal of Textile Universities
基金
国家自然科学基金青年项目(61901347)
大学生创新创业训练计划项目(S202010709104)。
关键词
服饰图像分割
着装人体
超像素
聚类分割
局部空间信息
模糊C均值算法
clothing image segmentation
dressed human body
superpixel
clustering segmentation
local spatial information
fuzzy C-means algorithm
作者简介
第一作者:李立瑶(1996—),女,西安工程大学硕士研究生;通信作者:顾梅花(1980—),女,西安工程大学副教授,研究方向为计算机视觉算法及其应用,E-mail:gumh2001@xpu.edu.cn。
