基于自适应尺度混合海森滤波器的面部皱纹检测及评分

Facial wrinkle detection and scoring mechanism based on adaptive scale hybrid Hessian filter

为解决全面部皱纹缺乏量化评价方法的问题,提出了一种基于自适应尺度混合海森滤波器(adaptive scale hybrid Hessian filter,ASHHF)的面部皱纹检测方法及评分机制。根据受试者年龄自适应调整海森滤波器的尺寸(σ)范围及步长参数并对其面部高分辨率图像进行滤波,基于81个面部特征点在滤波结果中去除五官、背景,仅保留面部皱纹,使用不同颜色将不同深度的皱纹检测结果标注在原始图像中,最后计算得分量化皮肤衰老程度。以专业医生标注皱纹为参考,50名受试者(年龄为22~65岁)的检测结果表明,相比于传统的固定尺度混合海森滤波器(fixed scale hybrid Hessian filter,FSHHF),ASHHF方法的检测准确率平均提升68.57%,运行时间平均缩短26.26%,评分机制结果与检测准确率的变化趋势一致。综上,本文检测方法能够准确、快速检测面部皱纹分布、深度及宽度,所提评分机制能够科学反映受试者皮肤衰老程度,有望为化妆品、医疗美容等抗衰方法提供功效量化评价手段。

A facial wrinkle detection method and scoring mechanism based on adaptive scale hybrid Hessian filter(ASHHF)are proposed to solve the problem of lacking quantitative evaluation method for the whole face wrinkles.According to the age of the subject,the size(σ)range and step parameters of the Hessian filter were adaptively adjusted,and the high-resolution facial images were filtered.Based on 81 facial feature points,the face and background features were removed from the filtering results,and only facial wrinkles were kept.The wrinkle detection results of different depths were marked in the original image with different colors.Finally,the score was calculated to quantify the degree of skin aging.Taking wrinkles marked by professional doctors as a reference,the detection results of 50 subjects(aged between 22 and 65)show that compared with the traditional fixed scale hybrid Hessian filter(FSHHF),the detection accuracy of the ASHHF algorithm was increased by 68.57%on average,and the running time was shortened by 26.26%on average.The result of scoring mechanism is consistent with the changing trend of detection accuracy.In summary,the detection method in this paper can accurately and quickly detect the distribution,depth and width information of facial wrinkles,and the proposed scoring mechanism can scientifically reflect the degree of skin aging of subjects.It is expected to provide a quantitative evaluation means for anti-aging methods such as cosmetics,medical cosmetology,etc.