사진.동영상 콘텐츠의 실시간 입체영상 변환 연구

Abstract

Humans’ eye disparity is 65mm or less. The binocular disparity is an important factor of stereoscopic effect. Humans’ stereoscopic feeling of an object is being caused by the double structure of eyes & rod(right brain) and nerve cone(left brain) organized by perceptual system. The rod of right brain affects to stare at an object, and the nerve cone of left brain makes possible to see an object in focused sight. In the principle of camera, they are similar with ‘out of focus’ and ‘pan focus’. The double structure of left & right brain does not works simultaneously, but it works by turns with delaying, by the double structure of brain such as above, humans are able to see the objects in stereoscopic 3D image. This perceptual system of human makes possible to express the image and video contents in stereoscopic by specific video devices, polarizing lens, red & blue glasses. The screen based stereoscopic images need captured images by two same cameras using the device which is called ‘Rig’. Except the stereoscopic 3D images such as pre-produced photos or film, video, animation contents are needed to be produced in frame by frame. For instance, to convert film duration of 100 minutes to stereoscopic contents in manual processes, it needs about 100 to 300 of edit crew and budgets. To complement the disadvantages such as these, in this paper, we suggest the method for converting existing 2D image contents to stereoscopic 3D image contents by real-time 2D/3D image conversion algorithm. The stereoscopic 3D image conversion is the method that is create a stereoscopic image from the photos or videos; and does not get the captured video from stereo camera. First, separate the objects and backgrounds from existing original 2D image, set the distancepixel value from the separated sources, then separate the image in left-right; then adjust the external parameter value of internal and external of camera. Then create a stereoscopic image by move the occlusion area and image according to the camera’s location. Suggested method above presents the three-dimensional effects without any limitations of scene changing; and it converts not only videos, but still images also to stereoscopic images in real-time. The real-time stereoscopic conversion algorithm which was organized above has produced based on the hardware which is called ‘Physical Board’ to complement the computational complexity from the limitation of the software’s algorithmic processes, and the drop of data processing speed. From the experiment, if we use the suggested method above, we can complement the stereoscopic 3D image contents by dimensionalize all of existing 2D image contents to 3D, moreover, we can also save budget for creating stereoscopic 3D image contents production.