2D Video Depth Perception Method Based On Indoor And Outdoor Scenes Classification

The demand for three-dimensional video sources is rapidly increasing for the significant decline in the price of the3D display devices. The technology of two-dimensional to three-dimensional video conversion has become the focus of attention, it can not only avoid the waste of resources but also reduce the cost. Depth extraction of2D image is an important research in the study of2D to3D conversion techniques. The main research of this paper is that high accuracy of indoor and outdoor scene classification can be obtained automatically for the input two-dimensional video, and different depth cues are adopted with the results of indoor and outdoor scene classification, then depth information of two-dimensional video sequence is obtained.The research status of depth extraction of two-dimensional video is overviewed on the basis of search of many domestic and international relevant research literatures.The methods of linear perspective and aerial perspective for depth extraction of two-dimensional videos are elaborated. The method that appropriate cue is selected to obtain the depth map according to different types of indoor and outdoor scenes is proposed.The relationship between the signal’s smoothness and Fourier spectrum and different features of indoor and outdoor scenes’smoothness is analyzed in the method of indoor and outdoor scenes classification. An effective method of indoor and outdoor scenes classification which is based on Fourier spectrum, color and texture features is proposed in this paper, and experimental results show that high accuracy of indoor and outdoor scenes classification can be achievedDifferent depth cues which adapt to different types of scene is analyzed after scene classification of the video sequence, the overall framework of2d video depth extraction based on indoor and outdoor scenes classification is proposed. Indoor and outdoor scenes classification can be achieved automatically by this framework, and appropriate cue is selected for depth extraction of indoor and outdoor scenes. The method of linear perspective and aerial perspective is utilized to extract depth of the indoor and outdoor sequences respectively. Experimental results have proven the effectiveness of proposed method.