在本篇研究中，我們提出一個人臉影像上的重建系統，給予系統一張測試影像，系統將會自動改變測試影像的光照條件。這個系統是困難的；因為光影分佈於人臉影像上的分佈區域與光源位置和被照射人臉三維幾何有關。為了解決這個問題，我們引用以學習為依據的方法，用訓練樣本影像來建立兩個光照條件之間的樣貌相關性。這樣的學習轉換式可以將人臉影像的光線分佈由輸入照明條一個條件(style Y)上。首先，本文方法利用Adaboost演算法分別建立分類器擷取光線和個人特徵的具鑑別性區域特徵，這些具鑑別特徵是揭露這兩個照明條件的主要光線與人臉特徵變異區域。這些鑑別性區域特徵進一步被用來幫助影像生成(光線生成過程），本文用的生成過程為馬爾可夫隨機場（MRF）方法。根據實驗結果來看，在MRF模型中使用了鑑別性區域特徵，能使得結果保留了系統輸入的個人特徵。 Abstract: In this study, we propose a system to relight face image. Given a test image, the system will automatically change its light condition. Such system is difficult since the dark and lighting areas of a particular lighting condition are highly depended on the lighting source position and the facial 3D geometry of the input subjet. To solve this problem, we introduce a learning-based approach to establish the appearance correlation between two lighting conditions in order to transform the lighting condition from the input lighting condition (style X) to another condition (style Y). First of all, the discriminative facial features of lighting and personal characteristics are selected, respectively, for verifying these two lighting conditions. The selection process is processed by Adaboost algorithm; these discriminative features are then used for guiding the relighting process (light reconstruction process). Such reconstruction process is based on the Markov random field (MRF) approach. According to the experimental results, the proposed system, where the discriminative features are integrated in the MRF graph model, makes the synthesis results preserve the personal characteristics of system input.
