Face Presence Estimation Based on Scene Recognition Approach

In this paper we investigate the possibility of human face presence estimation in images using a scene recognition technique as a multi-classifier method. In the first step, spatial pyramid matching algorithm is used for classifying images based on global and spatial information. Then, by using BoW approach for local features of the face and a linear support vector machine, the presence of a face in an image is detected. In fact, estimation of a face is done based on spatial and global information of image while local features of the face is not neglected. Important facial features of databases used for evaluation of the proposed method consist of different features of emotional facial expressions, facial details and, different lighting. The results demonstrate that the proposed method together with a pattern recognition approach could be a promising tool as an effective step for face detection field