基于双向预测插值的多维矢量鬼成像视频

为了进一步提升运动物体鬼成像视频的质量，根据视频中每帧都具有相关性的特点，提出了一种基于双向预测插值的多维矢量鬼成像视频的实时成像算法.即运用具有能量集中特点的多维矢量Walsh变换，将总数一定的Walsh基图样分成更少的组，使每组的有效基图样翻倍，再分别作用于目标物体，重构后得到数量少但成像质量更好的隔帧图像，再通过双向预测插值得到中间帧.使用该算法作用于像素为64×64的目标物体，可在实时状态下实现运动物体鬼成像的视频.在采样数量和采样时间不变的情况下，可实现高质量重构，得到了品质更好的鬼成像视频.

Multi-dimensional Vector Ghost Imaging Video Based on Bidirectional Predictive Interpolation

To further improve the quality of moving object ghost image videos， a real-time imaging algorithm based on bidirectional predictive interpolation for multi-dimensional vector ghost image video is proposed according to the correlation characteristics between each frame in the video. That is to say， the multi-dimensional vector Walsh transform with energy concentration is used to divide a certain number of multi-dimensional Walsh basis patterns into fewer groups， doubling the effective basis patterns of each group， and then acting on the target object respectively. After reconstruction， the number of interframe images with better imaging quality is less， and then the intermediate frames are obtained by bidirectional predictive interpolation. Using this algorithm， the ghost image video of moving object can be realized in real time by acting on the object with 64×64 pixels. With the same sampling quantity and sampling time， high-quality reconstruction can be achieved and better quality ghost imaging video can be obtained.