Research On Depth Enhancement And Rendering Techniques Based On Kinect

With the rapid development of digital video and the widespread application,describe the real world in 2d cannot satisfy people’s growing demand vision. The diversity demand people to video in content and quality is higher and higher, the 3d video visual more match with people, which can make people get rich stereo feeling and immersive. Multi-view stereo video can be provided to the audience more stereo video on, so relative to the current high-definition television, free view stereo video data volume is very big, is almost linear increase increased with the increase of number of viewpoints. However, the storage medium and the network condition for single-channel 2d high definition video applications leave much to be desired,certainly cannot meet the freedom viewpoints stereo video requirement for storage space and network bandwidth.if the multi-view stereo video want to popular among the people,the first thing should be to solve the problem of multi-view stereo video data. The key is to solve the data volume stereo video that representation of a 3d scene information problem, namely how to more effective representation of 3d scene to reduce data redundancy. This paper is mainly focused on the depth enhance Based on kinect and multi-view Image Based Rendering technology, and obtained the results as below:1.A novel method is put forward used to Update background model,during the process of the acquisition of depth image, use the initial frame to update the gaussian background model, in order to achieve the best background model matching.2.Put forward the enhancement algorithm based on gaussian mixture model,by using the gaussian mixture model(GMM) to separate the depth image of foreground and background,make the different treatment according to the holes appear area,the holes in background filling adopt corresponding value in the background,using texture map+ distance function + background of weighting function to determine the values in the foreground.3.Put forward the algorithm based ongeometric model of virtual perspective drawing method, The holes problems caused by Depth image based rendering(DIBR).First using the gaussian mixture model to separate the new point of view of the image to foreground and background, the holes be divided into foreground holes and background holes after DIBR,the holes in background use background information to fill and the holes in foreground with the improved image inpainting repair technology to deal with,At last a good new perspective images.