數位相片臉部過度亮點之偵測與修補

Automatic Detection and Correction Methods for Glossy Reflections in Digital Image

曾斐君(Tseng,Fei Chun)

指導教授 ： 張榮吉

亞洲大學/創意設計學院/資訊與設計學系碩士班/碩士(2009年)

摘要

數位攝影已成為人們記錄影像最便利的方法之一。當進行拍攝時，常常因為環境光的強度、光源角度及臉部本身的反光性，造成臉部區域出現過度曝光，產生亮點過大現象，而破壞了數位影像的美觀。本研究中，將針對數位相片臉部過度亮點的問題，以自動修補的方式提出有效的解決方法。本論文實驗中採用YCbCr色彩空間來界定膚色範圍，依據人臉特徵與人臉長寬比例做為偵測的篩選條件，找出人臉範圍後，針對人臉區域進行過度亮點偵測與修補。過度亮點偵測中會比較像素灰階值是否大於亮度閾值(Threshold)，若大於亮度閾值則標示為目標亮點，最後在目標亮點周圍尋找可用來源資訊，進行點對點修補避免造成影像模糊，讓修補後的影像在人類視覺感官上看起來更自然與美觀。

關鍵字

膚色辨識 ； 人臉偵測 ； 亮點偵測 ； 影像修補

並列摘要

Digital photography has become so popular in recent years. For most people, it is also a most convenient way of recording images. However, the intensity of ambient light often causes the highlight of facial region which making the serious destruction of digital photos. By using the automatic patching method, this paper aims to provide the effective solutions of removing the glossy reflections from human face. The proposed method employs the YCbCr color space to limit the skin color scope and to detect the glossy reflections area. The process of detecting the glossy reflection makes the precise comparison with the gray scale value of structuring element. If the average intensity is larger than the target threshold value, the detection will seek the available information around the target area to make the repair. Based on the results of this study, the digital photo becomes more natural and beautiful after repairing.

並列關鍵字

Skin Color Identification ； Face Detection ； Glossy Reflections ； Image Inpainting

參考文獻

[4]黃泰祥，「具備人臉追蹤與辨識功能的一個智慧型數位監視系統」，中原大學，碩士論文，民國93年

[2]林起字，「數位紅眼缺陷的全自動化檢測及校正研究」，大同大學，碩士論文，民國96年

[11] Safonov, Ilia V., “Automatic Red Eye Detection”, GraphiCon, pp.23-27, 2007.

[12] Criminisi, Antonio, “Region Filling and Object Removal by Exemplar-Based Image Inpainting”, Image Processing, vol.13, pp.1200 - 1212 , 2004.

[14] Gonzalez, Rafael C., et al., Digital Image Processing, Prentice Hall Press, Upper Saddle River, 2002.