A Depth Extraction Method Based On Fusion Of Inter And Intra Frame Cues Of Two-dimensional Video

3D movie market has become increasingly active in current, three-dimensional movies get more and more people’s favorite because that it can offer shocking feeling and visual experience to the audiences. However, there exist some problems, such as the cost of stereo camera filming and lack of3D resources,2D to3D video conversion technology has obtained widely attention. Depth extraction is the key technology in the process of two dimensional video to three dimensional video. As the limitations of depth extraction with one depth cue, this paper uses multi-cues to extract depth information of videos for better effect.Each depth cue has corresponding scene restriction through reading related home and abroad literature of two-dimensional video depth extraction technology. So for the inter frame information of videos, motion cue is used to extract video motion information; motion estimation algorithm based on block matching is used to obtain motion vector, and then depth information of inter frame is estimated. To intra frame information of videos, Texture and haze cues are used to extract depth feature of video, and then model training is utilized to estimate depth information of intra frame.On this basis, in order to obtain better effect of2D video depth extraction, this paper proposes depth extraction method of two-dimensional video based on cue fusion. The problem of motion region detection can be classified as sample classification problems in pattern recognition through calculating information entropy of adjacent video frames. The background region and motion region of video frame are two kinds of samples, the Neyman-Pearson criteria is used to determine the interface of samples and motion region can be achieved. Due to the depth information of background region in video frame is completely determined by the depth of intra frame, so depth of intra frame is adopted for background of the video frames; fusion function of inter and intra frame information is constructed for motion region of the video frames. Different mathematical functions are adopted to determine function of fusion coefficient, and then inter depth information and intra depth information can be fused effectively in order to improve the depth effect of2D video. Experimental results have proven the effectiveness of this method.