结合图像融合算法与人眼跟踪的立体视频

由于多视点立体视频合成具有数据量大，图像处理速度要求较高，支持的立体视角有限等特点，这些问题一直没有很好的解决并已成为多视点立体视频产业化的瓶颈。针对这种情况，提出了一种基于立体图像融合算法与人眼跟踪算法的立体视频处理系统。首先，按顺序循环读取立体视频中的每一帧，然后用立体图像融合算法对每一帧进行合成运算，接下来将融合后的图像依原有顺序进行显示与播放。同时加入人眼跟踪算法，根据观看者眼部所处的位置实时投放对应视区的图像。图像融合算法与人眼跟踪的结合有效地扩大了立体视角。实验结果表明，该方法实现了将多视点视频在立体显示器中以自由立体显示的方式展现出来，使观看者在屏幕前可以自由移动而不影响立体观看效果，同时播放速度流畅，能给观众带来比较真实的立体感受。

Multi-View Stereo Video Based on Fusion and Eye-Tracking

As multi-view stereo video has some new characteristics, such as massive volumes of data, high-demand in the velocity of transition, limited 3D view fields supported. These problems have not been well-solved and blocked the industrialization of multi-view stereo video technology. In view of this, this paper proposes a method of multi-view stereo video transition and display based on an image fusion algorithm and eye-tracking. First, we capture each frame from the original video, then, we use an image fusion algorithm, process each frame of the multi-view stereo video and output the frames at the original rate. At the same time, we use eye-tracking, which effectively enlarged the stereo view field. Experimental results show that the Two-dimension views image can be shown on the stereoscopic display in a 3D way. It could transmit the video smoothly and effectively while computing and eye-tracking. It could bring the viewers three-dimensional feelings when watching.