首页|3D沉浸式系统图像修复关键技术研究

3D沉浸式系统图像修复关键技术研究

扫码查看
对3D沉浸式系统图像修复关键技术进行了深入研究,提出了一种彩色图像引导的CNN压缩伪影抑制模型,对原始的非遗陶瓷文物图像的整体质量和边缘信息进行提升,增强修复效果.首先,对CNN卷积神经网络的基本结构进行了研究;然后搭建了基于CNN的深度图像压缩伪影抑制模型,同时引入谱分解概念对深度图像的高频部分进行重建操作,滤除无用的干扰信息,加快模型的训练收敛速度;并设计基于多尺度引导信息的内容修复方法增强了模型的修复效果;最后,对基于CNN的深度图像压缩伪影抑制模型进行了实验与测试.实验结果表明:引入了灰度信息辅助,即由彩色图像引导的CNN结构来搭建的模型性能,相比于没有引入灰度信息辅助的模型对图像的修复效果更佳;设计的模型在图像质量Q=30时,峰值信噪比为45.18,比其余5种实验对照组模型中性能最优的Liu模型增加了4.31;但图像质量Q=40时,峰值信噪比高达49.36,比Liu模型增加了 7.47,表明本文设计的图像修复模型性能更佳,图像修复效果得到显著提升,能够用于非遗陶瓷文物图像修复工作.
Research on Key Technologies of inpainting in 3D Immersive System
In this paper,the key technologies of inpainting of 3D immersive system are deeply studied,and a color image guided CNN structure is proposed to build a depth image compression artifact suppression model based on CNN,which improves the overall quality and edge information of the original intangible cultural relics image,and enhances the restoration effect.Firstly,the basic structure of CNN convolutional neural networks was studied;Then,a CNN based deep image compression artifact suppression model was constructed,and the concept of spectral decomposition was introduced to reconstruct the high-frequency part of the deep image,filtering out useless interference information and accelerating the training convergence speed of the model;And a content repair meth-od based on multi-scale guidance information was designed to enhance the repair effect of the model;Finally,experiments and tests were conducted on the CNN based deep image compression artifact suppression model.The experimental results show that the model constructed with a CNN structure guided by color images,which incorporates grayscale information assistance,performs better in im-age restoration compared to models without grayscale information assistance;The model designed in this article has a peak signal-to-noise ratio of 45.18 when the image quality Q=30,which is 4.31 more than the Liu model with the best neutral energy in the other five experimental control group models;However,when the image quality Q=40,the peak signal to noise ratio is as high as 49.36,an increase of 7.47 over Liu's model,indicating that the inpainting model designed in this paper has better performance,and the in-painting effect has been significantly improved,which can be used for inpainting of intangible cultural relics.

3D immersive systeminpaintingintangible cultural heritage ceramic relicsCNN

姚翊姁

展开 >

长沙师范学院,长沙 410000

3D沉浸式系统 图像修复 非遗陶瓷文物 CNN

湖南省教育厅科研项目

20C0129

2024

自动化与仪器仪表
重庆工业自动化仪表研究所,重庆市自动化与仪器仪表学会

自动化与仪器仪表

CSTPCD
影响因子:0.327
ISSN:1001-9227
年,卷(期):2024.(1)
  • 14