基于互反射先验的端到端沉浸式投影补偿  

End-to-end projection compensation in immersive projection systems based on the inter-reflection prior

作　　者：胡宗禹 程鹏 刘建 刘洪[1] 陈虎[1,4] 张意 杨红雨[1] HU Zong-Yu;CHENG Peng;LIU Jian;LIU Hong;CHEN Hu;ZHANG Yi;YANG Hong-Yu(College of Computer Science,Sichuan University,Chengdu 610065,China;School of Aeronautics and Astronautics,Sichuan University,Chengdu 610065,China;Airspace Management Center,Air Traffic Management Bureau,Beijing 100000,China;State Key Laboratory of Fundamental Science on Synthetic Vision,Sichuan University,Chengdu 610065,China;School of Cyber Science and Engineering,Sichuan University,Chengdu 610065,China)

机构地区：[1]四川大学计算机学院,成都610065 [2]四川大学空天科学与工程学院,成都610065 [3]中国民用航空局空中交通管理局空域管理中心,北京100000 [4]四川大学视觉合成图形图像技术重点学科实验室,成都610065 [5]四川大学网络空间安全学院,成都610065

出　　处：《四川大学学报（自然科学版）》2024年第1期51-60,共10页Journal of Sichuan University(Natural Science Edition)

基　　金：国家自然科学基金民航联合研究基金(U1833128);成都市重点研发支撑计划项目(2021YF0501788SN)。

摘　　要：沉浸式投影系统中的互反射现象会导致投影质量下降,投影补偿任务通过对输入图像的修改补偿,减少互反射造成的影响从而得到更好的成像效果.投影补偿任务是研究输入原图与最终成像之间的关系,通过对输入图像的修改补偿,减少互反射造成的影响从而得到更好的成像效果.本文提出互反射先验,互反射先验是指发生互反射的区域的临近区域通道像素值十分接近最大值256,因此通过图像的区域通道像素值可以获得互反射信息.本文提出的互反射补偿网络(IRCN),通过互反射先验生成互反射掩膜,消除沉浸式投影环境中的多余互反射.为了验证IRCN模型,使用了超过5000对投影图像数据集和曲面投影系统以及折面投影系统来完成实验.本文将IRCN对比多种现有投影补偿方法,实验结果表明IRCN在PSNR,RMSE,SSIM,均值标准差分析等客观评价指标中具有明显优势,说明本文提出的IRCN能够有效利用互反射先验信息,对沉浸式投影环境中的互反射现象进行补偿,在均衡训练时间的基础上增强了投影效果.The inter-reflection phenomenon in immersion projection systems seriously affects the effect of projection,the projection compensation task can reduce the influence caused by the inter-reflection by modifying and compensating the input image so as to obtain a better imaging effect.In this paper,an inter-reflection prior is proposed,which means that the color effect of the region where inter-reflection occurs is very close to the maximum 256 due to the light superposition effect.Therefore,inter-reflection information can be obtained through the regional channel pixel of the image.The Inter-Reflection Compensation Network(IRCN)proposed in this paper generates an inter-reflection mask value through the inter-reflection prior,so that the IRCN can eliminates the excess interreflection in immersive projection environments.To verifly the IRCN model,the erperiment were performed on more than 5000 pairs of projected image datasets and curved projection system,and fold projection system.Compared with the existing projection compensation methods,the experiment results show that IRCN has obvious advantages in objective evaluation indicators such as PSNR,RMSE,SSIM,mean and standard deviation analysis,The proposed IRCN can effectively utilize the inter-reflection prior information to compensate for the inter-reflection phenomenon in the immersive projection environment,and enhance the projection effect based on balancing the training time.

关 键 词：投影补偿 神经网络 互反射 沉浸式投影 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]