Research And Implementation Of Multi-View Stereoscopic Video Systems

With the development of multimedia and network communication technology, video, as the most important modality of multimedia, has been widely used in industrial production and people’s lives. Although the traditional high-quality two-dimensional video can display the texture of an object, it cannot bring people the perception of the depth. In order to make it easier for people to watch high-quality three-dimensional videos, the new generation multi-view stereoscopic video system emerges now. Usually a typical multi-view stereoscopic video system consists of capture module, calibration module, coding module, transmission module, decoding module, and three-dimensional display module. In this thesis, we study the principle of multi-view stereoscopic video system, make a number of innovations and implement the system. The work mainly includes the following aspects:(1)The videos captured by multi-view cameras exist horizontal and vertical mismatches as well as scaling and rotation mismatches. In order to improve the display quality of the final synthesized three-dimensional videos, an improved multi-view video calibration algorithm is proposed;(2)The calibrated multi-view videos will be encoded and transmitted. In order to reduce the huge amount of data, an optimized intra coding and sampling algorithm for nine-viewpoint format video are proposed;(3)The decoded multi-view videos will be synthesized before they are displayed on a naked eye three-dimensional display device. An algorithm of selecting zero disparity point which are controllable by parameters is proposed;(4)Based on the algorithms above, a multi-view stereoscopic video demo system is developed.