基于CycleGAN的4K超高清电视色域及动态范围转换算法研究

Research on Color Gamut and Dynamic Range Conversion Algorithm for 4K UHDTV Based on CycleGAN

本文在研究超高清电视色域、动态范围技术的基础上,首先拍摄制作了4K超高清视频序列形成训练数据集,随后提出了一种改进的基于循环生成对抗网络(CycleGAN,Cycle Generative Adversarial Network)的深度学习网络,通过研究及多次试验,确定了损失函数。最终通过网络训练实现了基于CycleGAN的4K超高清电视色域、动态范围之间的转换,并进行了效果分析。

Based on the research of UHDTV color gamut and dynamic range technology,this paper first shots and produces 4 K UHD video sequences to form a training data set,and then proposes an improved deep learning network based on CycleGAN.Through research and many experiments,loss function is determined..Finally,through network training,conversion between color gamut and dynamic range of 4 K UHD TV based on CycleGAN is implemented,and effect is analyzed.