- 學位論文
數位紅眼缺陷的全自動化檢測及校正研究
AUTOMATIC RED-EYE DETECTION AND CORRECTION IN DIGITAL PHOTOGRAPHS
摘要
紅眼是在使用數位相機拍攝時一個很令人困擾的問題，雖然迄今已有許多消除紅眼的演算法被提出且實現於數位相機中，但效果皆相當有限。在本篇論文中，我們提出一個全自動的紅眼缺陷檢測以及校正的演算法，期望在不同拍攝角度下產生的紅眼，都能夠被正確地檢測出來，並將該被檢測出來的紅眼恢復成在視覺上自然之效果。 為了檢測紅眼，首先在 色度空間下，將影像分割出紅色以及膚色的區域；接下來利用紅眼濾波器將紅色的區域濾除成紅眼候選區域，並對每一紅眼候選區域計算其弧度來判別是否為紅眼；接著驗證紅眼區域是否包含在膚色區域內，以判定正確的紅眼定位；最後藉由調整像素上的色彩飽和度及亮度來校正我們所檢測出之紅眼，使其在視覺感受上能夠較為自然。 我們收集了150張具有紅眼的相片來評估所提演算法的效能。實驗顯示，超過85%的紅眼可以被正確地檢測並消除，此結果證明我們所提出來的方法是強健而且有效的。
並列摘要
Red-eye is a troublesome problem in digital photographs. Although many red-eye reduction algorithms were proposed and carried out in most of the digital cameras, none of these algorithms is effective enough. In this thesis, we propose an automatic red-eye detection and correction algorithm to improve previous work. Multiple pairs of red-eyes snapped in different view angles are expected to be detected correctly and recovered to be a natural look in digital photographs. In order to detect red-eyes, an image with red-eyes is first divided into the red color regions and skin color regions in the color space. Then we use a red-eye filter to sift out red color regions and denote them as red-eye candidates. After obtaining red-eye candidates, we calculate the radian of each red-eye candidate to verify if it is included in the skin color regions and subsequently determine its location precisely. Finally the color of the detected red-eyes is corrected by modifying the saturation and luminance of the associated pixels such that red color is removed while maintaining a natural look. We evaluate the performance of the proposed algorithm with a collection of about 150 red-eye images. Experimental results show that more than 85% of red-eyes can be detected, thus suggest that the proposed method is rather robust and efficient.