皮肤老化分析中的皮沟检测方法研究<br>Research on Detecting Skin Grooves in Skin Aging Analysis

Abstract

皮沟检测是客观量化人体皮肤老化程度的一个关键问题，检测精度直接影响到后续皮肤表面二维几何特征的计算。本文提出了一种基于数据驱动和模型驱动的混合控制策略来检测皮沟。其中数据驱动的控制先对输入图像进行一系列增强处理，再进行分水岭变换以获得粗略的检测结果；而模型驱动的控制则利用数据驱动处理的结果及视觉感知的先验知识建立一个区域合并模型，该模型被用于移除粗略检测结果中的虚假皮沟。主观和客观实验表明，对不同粒度的皮肤图像，提出的方法均能精确地检测出皮沟，并能有效地抑制虚假皮沟的出现。 Detecting skin grooves plays an important role in the objective quantification of human skin aging, and the detection accuracy directly influences the subsequent computation of two-dimensional geometrical characteristics of the skin surface. Based on the combined control strategies consisting of data-driven and model-driven control, a novel seg-mentation approach to skin grooves detection was proposed in this paper. In data-driven control, the rough segmentation result was obtained by applying watershed transform to the enhanced image. In model-driven control, the information acquired from the data-driven control and prior knowledge derived from sense of vision were used for constructing a region merging model. The model was used for removing redundant watershed lines. Subjective and objective evalua-tions demonstrate the favorable performance of the proposed method in detecting precisely skin grooves and restraining false skin grooves.