Pixel-Based Hierarchical-Feature Face Detection

In this paper, the Pixel-Based Hierarchical-Feature Adaboosting (PBHFA) method is presented. The purpose of this approach is the reduction of computation complexity in face-detection tasks. The Adaboosting method has attracted attention for its efficient face-detection performance. However, in the training process, the large number of possible Haar-like features in a standard sub-window becomes time consuming, which makes specific environment feature adaptation extremely difficult. For this object, the PBHFA is proposed as a possible solution. Given a M x N sub-window, the number of possible PBH features is simplified down to a level less than M x N, which significantly reduces the length of the training period by a factor of 1500. Moreover, when the trained PBH features are employed for practical face-detection tasks, the hierarchically structural pattern matching also has lower complexity than that of the integral-image based approach in the traditional Adaboosting method. As documented in experimental results, with the MIT-CMU profile test set are examined, the proposed PBH features have shown significantly more effective than Haar-like features.