摘要
面部毛孔特征是人脸识别、皮肤检测的重要指标之一.为了有效地消除其他皮肤特征对毛孔检测的影响,提出一种融合皮肤色素分布特性及最佳尺度的高精度毛孔检测算法.针对皮肤特征在不同色素层上的色素分布特性相异,首先根据皮肤特征显著性差异和K-means聚类为SURF和SIFT算法设置合理阈值检测不同色素层图像上的皮肤特征;然后引入欧氏距离描述不同色素层检测点的位置信息相似性,利用最佳尺度作为阈值有效地筛除非毛孔特征干扰项.在毛孔检测的基础上,利用SIFT算法中的最佳尺度构建了皮肤毛孔粗糙度评价指标.选取Bosphorus人脸库中的正脸图像进行毛孔检测及评价的对比实验,结果表明该算法提高了毛孔检测的准确度,构建的评价指标稳定有效.
Facial pore feature is one of important indicators for face recognition and skin defects detection. In order to eliminate the effect of other skin features from facial pore detection processing effectively, we proposed a new facial pore detection algorithm that coalesces the characteristics of skin pigment distribution and optimal scale. Considering the dissimilarity of skin pigment distribution on different pigment layers, reasonable thresholds were set for SURF and SIFT algorithms to detect the skin features on different pigment layers, which was based on the significant difference of skin features and K-means clustering method. Then, the Euclidean distance was calculated to describe the positions similarity of the same detected points on different layers. Last, the optimal scales were set as thresholds to screen off the interferences of skin features except pore. On this basis, an evaluation criterion of skin pores roughness was also established by using the optimal scales obtained during SIFT process. We selected positive face images from Bosphorus dataset for comparative experiments about pore detection and evaluation. The experiment results show that the proposed algorithm improves accuracy of pore detection, and the evaluation criterion is stable and effective.
作者
华斌
李若瑄
王朕
Hua Bin;Li Ruoxuan;Wang Zhen(School of Science and Technology,Tianjin University of Finance and Economics,Tianjin 300222)
出处
《计算机辅助设计与图形学学报》
EI
CSCD
北大核心
2019年第6期951-960,共10页
Journal of Computer-Aided Design & Computer Graphics
基金
国家自然科学基金青年科学基金(61502331)
天津市自然科学基金(16JCYBJC42000,18JCYBJC85100)
关键词
图像处理
毛孔检测
色素分布
最佳尺度
毛孔粗糙度
image processing
pore detection
pigment distribution
optimal scale
degree of pore roughness
