高动态范围(HDR)立体视频具有更高的亮度、更大的位深以及更广的色域，越来越受到人们的欢迎。然而，随着网络的发展，HDR立体视频可能被非法利用，其版权受到严重的威胁。现有显示设备都为低动态范围(LDR)，HDR立体视频必须通过色调映射，显示在这些设备上。因此，不同于LDR视频，HDR立体视频水印既能够保证HDR立体视频与其色调映射后的视觉质量，又能够有效保护它们的版权。本项目具体内容为：（1）将从主观感知探索人眼关于HDR立体视频失真对其本身以及LDR立体显示的视觉感知影响，挖掘视觉掩蔽规律并建立用于水印指导的视觉感知模型；（2）结合挖掘的感知因素，采用深度学习等方式建立HDR立体视频的HDR立体恰可察觉失真模型；（3）挖掘HDR立体视频的视点内相关性、视点间相关性、时域相关性等特征，提取鲁棒特征域，并结合视觉感知模型设计水印模型。本项目将为HDR立体视频系统的安全性提供理论思路与方法。

High Dynamic Range (HDR) stereo video is more and more popular because of its higher brightness, larger bit depth and wider gamut. However, with the development of the network, HDR stereo video may be illegally used, and its copyright is seriously threatened. Since existing display devices are low dynamic range (LDR), HDR stereo video must be displayed on these devices through tone mapping. Therefore, unlike LDR video, HDR stereo video watermarking can not only guarantees the visual quality of HDR stereo video and its tone mapped video, but also effectively protects their copyright. The specific contents of this project are listed as follows. (1) The visual perception effect of HDR stereo video and LDR display is explored from subjective perception because of distortion of HDR stereo video, and rules of the visual masking will be excavated and established for watermarking. (2) Combined with the perception factors, distortion model of HDR stereo video will be established by using deep learning. (3) The features of intra-view correlation, inter-view correlation and time-domain correlation of HDR stereo video are mined to extract the robust feature domain, and the watermarking model based on visual perception model is designed. This project provides theoretical ideas and methods for the security of HDR stereo video system.