Volume 2 Issue 2
Authors: Xiaohu Ruan; Zhen Tian; Tongyu Li; Hong Qin; Weijun Li
Abstract: Facial gender recognition is a strict application of computer intelligence. The traditional researches on gender recognition usually utilize a certain database, in which the face images of the training set and test set are taken under the same condition and from the same experimental population. That definitely cannot meet the requirement for practical application. In this paper, a novel experiment method is proposed by establishing a new Face Database to simulate real application; and also a facial gender recognition method is proposed. The facial gender recognition method employed Local Gabor Binary Pattern (LGBP), Principal Component Analysis (PCA) and Orthogonal Linear Discriminant Analysis (OLDA) for feature extraction, and a K-Nearest Neighbour (KNN) voting strategy for gender judgement. Experimental results showed that the robustness and generalization ability of this method are better than those of traditional methods. It provides a new perspective for gender recognition.
Keywords: Gender recognition; Face image; Local Gabor Binary Pattern; Orthogonal Linear Discriminant Analysis
Authors: Zhichao Yang; Guanglin Yang; Haiyan Xie
Abstract: A naked-eye three-dimensional (3D) video real-time display system is proposed. In this system, the cameras number of capturing 3D information were reduced from nine to two, and the naked-eye 3D images multi-view display was realized using the 2 cameras. The disparity image and the left image from stereo views were applied to generate the intermediate seven virtual viewpoint images along the horizontal baseline. The Adaptive Support-Weight (ASW) algorithm with Variable Support Window (VSW) was applied to obtain the disparity image. The depth-image-based rendering and image interlacing were adopted to get a multi-view image (i.e. 3D information). The screen of integral-imaging using lenticular lens was applied to display 3D images with continuous and wider viewing zone. The high speeding parallel programming architecture CUDA was applied to realize the high-quality naked-eye 3D video real-time display. Experimental results showed that the stereo-matching algorithm was suitable for the proposed system.
Keywords: Naked-eye 3D; Virtual Viewpoint Images; Image-based Rendering; Real-time Display