Research On Depth Generation For 3D Scenes

The obtaining accurate depth is one of the key technologies in these 3D applications and 3D researches.With the attractive advance of 3D technology,many applications,such as: human-computer interaction,3DTV,3D reconstruction,gesture recognition,augmented reality demand a higher requirement for scene’s depth information.These applications need more dense and accurate depth information,and need a faster acquirement for depth information.This paper elaborates two ways to depth generation: passive range sensors and active range sensors.Based on sufficient survey of two ways,this paper makes some improvements on the existing stereo matching algorithm,up-samples sparsely depth map of TOF,and makes the depth integration which combined with the passive stereo matching algorithm.The major innovations and contributions of this paper are described as follows:1 Aiming at mistakes of traditional stereo matching algorithms,we propose a novel method which includes stereo matching in the pixel level and surface interpolation in the image level,experiment results show that it outperforms all the other local methods using edge-aware filtering in terms of accuracy and speed.2 This paper studies the optical imaging system of TOF depth camera,and de-noises the depth map obtained by TOF.Due to the resolution of the depth map acquired by TOF depth camera is limited,the depth map does not match with the high-definition color image,and in order to avoid adding personal error,so TOF depth map must be up-sampling sparsely for higher resolution.3 In order to obtain the high accurate final dense depth map,this paper shows a novel TOF-stereo fusion method which estimates depth using TOF in weakly/ repetitively textured scenes and stereo matching in highly textured scenes as much as possible.Experiment results show that our fusion algorithm improves the accuracy and speed,the generated depth map is much better than that obtained from an individual method.