人臉特徵偵測與分析

Abstract

本論文提出一個基於灰階閥值和色域範圍的人臉特徵偵測方法。在本論文中，我們採用Canny邊緣偵測、自適閥值、CIELAB色彩空間、HSV色彩空間等技術，對高解析度的人臉影像進行皮膚紋理、皺紋、斑點、毛孔、青春痘、敏感、膚色均勻程度之特徵提取，透過與一般皮膚區域面積的比例判斷各種症狀的嚴重程度並給予相對應的評價。

A method of human face feature detection and analysis based on gray-scale threshold and color space value range is proposed in this thesis. In our work, we use techniques such as Canny edge detection, adaptive threshold, CIELAB color space, HSV color space, and so on to extract features including fine lines, wrinkles, spots, pores, acnes, sensitivity, and skin color uniformity from high-resolution human face images. We also provide the severity level of each symptom by measuring the area ratio of extracted features to normal skin and give corresponding evaluation.