Study Of Multi-view Coding And Object Segmentation Technique Based On MPEG-2

As on DTV / HDTV two-dimensional (2D) video after the next generation of video technology, three-dimensional (3D) stereoscopic video has become a hot topic in international research. Compared to single-channel video and two-dimensional, an increase of three-dimensional video scene depth information, and enhance the visual realism and a sense of reality. View more view video increased the number of cameras by using multiple parallel or convergent arrangement of the same scenes multiple perspective image, the viewer can select viewing angle, any perspective of a different image. Multi-view video applications have limitations, first of all is that it's coded transmission problems. As many times in the number of binocular stereo video point of view, is correspondingly doubled the amount of data, how to become a multi-view video coding efficiency of multi-view video of the current hot spot.At present, a variety of multi-view coding schemes, but summed up is the use of motion estimation and compensation techniques to eliminate temporal redundancy, the use of disparity estimation and compensation techniques to eliminate inter-view redundancy. The one hand, these advanced coding idea both advantages and disadvantages: on the one hand, it allows applications to achieve multi-view video system possible, on the other hand is too computationally intensive to bring a large number of lossy compression, image quality poor.In this dissertation, single-frame "8-Title" with the "MPEG-2" General two-dimensional multi-view video coding standard encoding scheme. Numerous and diverse compared to encoding and decoding multi-view programs, single-channel video compression standards such as MPEG-2, regardless of technical specifications, or product development are mature, the implementation section of this paper is the use of this coding standard.The dissertation also launches one hot research of the current three-dimensional video (2D turn 3D). Based SFM technology, 2D to 3D conversion system automatically detects scene changes by early to estimate the static background and moving foreground objects segmentation independent, the method can be applied to the depth of the image assignment.