自動眼睛偵測及眼鏡去除方法

Abstract

在本篇論文中，我們提出一演算法，使其能夠在一張人臉影像中偵測眼睛的位置，並且在該人臉有戴眼鏡時，去除眼鏡的干擾。此系統包含三個模組：臉部位置偵測，眼睛偵測，以及當受測者有戴眼鏡時，去除眼鏡的方法。首先，我們使用通用膚色圖偵測臉部區域，以確保在周遭光源條件變化時，仍有足夠的適應性。接著，採用在臉部傾向及旋轉上有良好不變性的環狀頻率濾波器，偵測眼睛區域。最後，由於眼鏡的普遍使用，提出以邊緣偵測及改良的模糊規則型式的濾波器，預測眼鏡區域的像素值。結果顯示，我們提出的方法在眼睛偵測是有效率的，並且在眼鏡去除上也能得到良好的品質。

This thesis addresses an algorithm to automatically detect the eye location from a given face image and remove the glasses while one has worn. Our system consists of three modules: face segmentation, eye detection, and eyeglasses removal while one has worn on eyeglasses. First, we use the universal skin-color map to detect the face regions, which can ensure sufficient adaptability to ambient lighting conditions. Then, a special filter, called circle-frequency filter, is used to locate the eye regions because of its invariant characteristic in wide face orientations and rotations. Finally, for the widely using of glasses, we proposed a novel method to remove the eyeglasses automatically based on edge detection and modified fuzzy rule-based (MFRB) filter. The simulation and results demonstrate that our approach detects eye location efficiently and demonstrates high fidelity to the non-wearing-glasses facial image after glasses removal.