Dimensional 3D Game, 2D Video To Stereo 3D Technology Research

"3D", this word has been given another meaning with the development of times. Nowadays the meaning of the word has been talked about the most is the "3D" based on the principle of binocular stereoscopic vision. With the increasing popularity of this 3D technology 2D image has not been satisfied people. The watching requirements of people are developing to 3D image that the depth can be felt from. But a very realistic problem is that the 3D resourses are too few to meet the requirements of people. So 2D to 3D technology which can convert 2D recourses to 3D resources will be a good solution in the interim of "2D to 3D"2D to 3D technology which been realized in ASIC chip has been used in 3DTV in most cases nowadays. This technology is not mature and the effect of conversion can not be compared with the one which is given by 3D photographic equipments. In industry this technology has been called "pseudo-3D". Only POWERE DVD and Tridef 3D Experience video player have the function of 2D to 3D in software. Their effect of conversion is too poor to see the 3D effect as a result of sample conversion algorithm. The reason why software which has the function of 2D to 3D conversion is so few and the effect of conversion is so poor is that the conversion algorithm is so complicated and imperfect that the CPU can not deal with video stream with the conversion algorithm in real time.This article will focus on realizing the conversion of plane 3D vedio games to stereoscopic 3D ones by C++ programming. Only a key technology which is edge detection in 2D vdeio stream to 3D one will be realized as a result of time limitation of project. This project will use GPU which has much power in performance but has been ignored by people and CUDA which is provided by NVIDIA Company to finish.Firstly this article will make a detailed introduction about principle of binocular stereoscopic vision.Next the project, will realized the conversion of the conversion of plane 3D vedio games to stereoscopic 3D ones. Plane 3D vedio games have 3D features in nature. So we just need to extract depth information from 3D vedio games without image processing and use Depth Image-Based Rendering to finish anaglyphic 3D conversion. In this project we still need to use DLL injection and API interception to change the way of program running for 3D conversion. Objects for conversion is 3D vedio game or other 3D program which made by directx9 API which is provided by Microsoft.Then the article will make a sample introduction about General-purpose computing on GPU and CUDA workbench.At last this article will introduce edge detection and object extraction which are key technology in this poject. All algorithms will be excute by GPU in this poject. All objects which will be chosen for testing are movie resources from internet. We will use improved canny operator to finish edge detection and use edge connection algorithm to make a good edge map which has complete and continuous edge lines. Complete and exact edge map will be very useful to realize image segmentation and depth estimation later. In object extraction we will use a new method called "four setp". Then we will produce a depth map and use DIBR to finish the final anaglyphic 3D conversion.