Researches On Depth Compression In 3d Video

Three dimension video (3DV) is an emerging multimedia application in our life, and is still under rapid development. Compared with traditional two dimension video, the new 3DV can provide the users with the depth information of the object, which can lead user to a stereo world. The current 3D technology allows the user to select a specific view point and switch view angle. We can expect a wild application market for 3D technology, e.g. 3D camera, 3D conference, 3D satellite image, human computer interaction. 3D video has a huge size as it contains the multi-view texture videos and the corresponding depth videos. Whether the 3D technology can be widely applied depends on whether we can compress the 3D video effectively. Fortunately 3D video has some new features, such as motion relevance between texture and depth, no detail in depth, which inspire us to deal with the problems in novel ways. Consequently, researches on 3D video are of great value in theory and practice.As we mentioned above, 3D video contains multi-view texture videos and their corresponding depth videos, based on such combination manner, this paper mainly works in two aspects to deal with depth information. The main content and novelties are listed as below:1. A new 3D video coding method ultilizing the character of depth is proposed in this paper.Texture video and depth video is the basic component when building a stereo world. However some research works reveal that the human visual system is more senstive to texture rather than depth when scene cut happens. A new coding algorithm is therefore proposed to compress 3D video, where technology Layered Depth Video and Multiview video coding are employed. Simulation results show that a compression gain can be achieved with half of the tested sequences. A subjective research was made to show that the new method would not bring down the subjective quality.2. Research on the roles depth video plays when synthesizing virtual viewIt is very meanful to make some researches on revealling the role depth plays when synthesizing virtual view. Depth video presents in the same format with the texture video and has big size. However user would not oberve depth videos, which only work in the backstage. Some research works have already shown that quality of the vitual view QualityV irtualViewdoes not necessarily depends on the quality of depth Quality depth. And if the quality gets down, virtual view would not go bad correspondingly. However there is still no definitely function established to demostrate the relation between QualityV irtualView and Quality depth.A hypothesis is proposed and proved in this thesis that in 3D video application systems, depth can be allocated with low resolution, which is sufficient for view synthesis.