欢迎您！东篱公司
退出
-
申报数据库
-
立项数据库
-
成果数据库
-
项目获奖数据库
中文摘要：
在自由视点视频系统中，由深度估计算法得到的深度视频并不精确，对虚拟视点的绘制质量和深度视频的编码效率有很大的影响。本文提出了一种基于自适应窗口的深度视频预处理算法来提高深度视频的压缩效率和虚拟视点的绘制质量。首先，将深度视频划分为不连续区域、连续区域中的前景区域和连续区域中的背景区域。然后，提取出深度视频各区域中的对象边界。最后对不连续区域中的非边界区域进行高斯滤波，降低深度值骤变对虚拟视点绘制质量的影响，对连续区域中的非边界区域采用自适应窗口滤波平滑处理以提高压缩效率。实验结果表明，提出的算法可以使深度视频的编码码率节省8．33～34．39％，同时可使绘制的虚拟视点质量平均提高0．21dB。和Silva的算法相比，在更具实际意义的低码率端编码效率更显著，绘制结果中对象边界得到了更好的保护。
英文摘要：
In free viewpoint video （FVV） system, depth video is the geometric information of three dimensional scene. It is utilized to render the virtual views. The depth video obtained by the depth estimation algorithm is not accurate, which deteriorates the quality of virtual views and degrades the compression performance. A depth video preprocessing algorithm based on adaptive window is proposed in this paper to improve depth video compression efficiency and rendered virtual views quality. Firstly, depth video is classified into discontinuous region, foreground region in continuous region and background region in continuous region. Then,object edges are extracted in all regions of the depth video. Finally, the non-edge part in discontinuous region is smoothened by Gaussian filter to reduce the impact of sharp depth transition on virtual view rendering. The non-edge part in continuous region is smoothened by adaptive window filter for high compression efficiency. Experimental results show that the proposed algo- rithm can save encoding bit rate ranging from 8. 33% to 34. 39% ,while it improves the quality of the rendered virtual views by 0. 21 dB on average. Compared with Silvars algorithm,the proposed algorithm achieves higher compression performance at low bit rate which is significant for depth video encoding, and keeps better object boundaries in the rendered virtual views.
同期刊论文项目
同项目期刊论文
Stereoscopic Visual Attention Based Regional Bit Allocation Optimization for Multi-view Video Coding
Colour Correction Preprocessing and Chrominance Reconstruction Postprocessing for Multi-view Video C
New Stereoscopic Image watermarking Method Based on Correlation of Inter-view Block Pair for 3D Medi
Color Correction Preprocessing and Chrominance Reconstruction Postprocessing for Multi-view Video Co
期刊信息
位置：