基于深度学习的投影互反射补偿

Inter-Reflection Compensation Based on Deep Learning

沉浸式投影系统的互反射补偿是虚拟现实技术中常被忽略,但又非常重要的研究课题。在沉浸式投影系统中,常常因为屏幕和投影仪之间的光线互反射造成投影图片质量严重下降,从而影响虚拟现实系统的真实性。现有的方法采用求解光传输矩阵的方式消除反射光,存在矩阵求逆复杂,效果不佳等问题。采用深度学习对投影图片进行补偿,能够避免复杂的矩阵求解过程,具有更优的视觉效果和ROI表现。

Inter-reflection compensation of immersion projection system is often ignored in virtual reality technology,but it is also a very important re⁃search topic.In immersive projection system,the quality of projection image is seriously degraded due to the mutual reflection of light be⁃tween screen and projector,which affects the authenticity of virtual reality system.The existing methods eliminate the reflected light by solving the optical transmission matrix,which has the problems of complex matrix inversion and poor effect.Using deep learning to compen⁃sate the projection image can avoid the complex matrix solving process,and has better visual effect and ROI performance.