Video 2D To 3D Technology Research And Implementation For 3DTV

3D movies aggressively promote the development of 3DTV industry.3DTV industry has a broad business outlook. Long shooting period, high production costs of 3D television works led to the shortage of 3D Video resources, restricting the development of 3DTV industry. A large number of existing 2D video can be converted into 3D formart by vedio 2D to 3D technology. This technology can not only enrich the existing 3D video resources, but also improve the existing 2D video viewing experience.Puts forward the vedio 2D to 3D solution, based on the principles of binocular stereo vision. The thesis presents a 2D video depth extraction method based on dense optical flow and modifies Gaussian pyramid transform and Laplace pyramid transform to fill the holes in the virtual viewpoint image. The procedure is as follows:(1) Using the motion parallax depth cues in 2D video to extract depth information, dense optical flow algorithm is used to extract the dense optical flow motion vectors in the video, then the dense optical flow motion vectors are transformed into depth information, finally the depth information is refined to obtain a final depth information that edge contour of the objects in the scene maintain good.(2) Using the depth information and vedio color image, virtual viewpoint synthesis technology based on Shift-Sensor camera model, to synthesis a new virtual viewpoint image. Synthetic virtual viewpoint image and the original color images constituting a stereoscopic image having binocular parallax relationship. Gaussian pyramid transform and Laplace pyramid transform are modified in this thesis to fill the holes of the synthetic virtual viewpoint image.(3) Performing experiments on the standard dataset and comparing results with the current mainstream algorithm to verify the validity of the algorithm.Experimental results show that the scheme can extract depth information that edge contour of the objects in the scene maintain good from the 2D vedio. Using the extracted depth information combined with video color image, virtual viewpoint synthesis technology based on Shift-Sensor camera model, can synthesize a virtual viewpoint image. Modified pyramid transform can fill holes in the synthetic virtual viewpoint image, reduce the viewer's discomfort.