基于场景运动程度的深度视频时域一致性增强
发布时间：2018-11-13 09:53
【摘要】：深度视频静止区域普遍存在深度值时域不一致,导致编码效率下降且影响绘制质量。针对该问题,提出一种基于场景运动程度的深度视频一致性增强算法。首先,应用基于块的直方图差值(Block Histogram Difference,BH)对深度视频每帧之间做相对运动程度量化度量,根据BH值自适应选取运动程度相对最弱的视频段作为深度值修正源,通过运动检测对相应彩色视频做运动区域分割,接着,利用彩色视频准确的时域一致信息,对深度视频中静止区域的错误变化深度值进行时域一致性校正,最后应用计算复杂度低的时域加权滤波函数对校正后的深度视频进一步优化,得到时域一致性优化的深度视频。本文算法相比于原估计获得的深度视频节省编码码率17.48%~31.75%,深度图所绘制的虚拟视点主观质量提高。
[Abstract]:The inconsistency of depth time domain exists in the static region of depth video, which leads to the decrease of coding efficiency and affects the quality of rendering. To solve this problem, a depth video consistency enhancement algorithm based on scene motion degree is proposed. Firstly, a block-based histogram difference (Block Histogram Difference,BH) is used to measure the relative motion degree of the depth video between frames. According to the BH value, the video segment with the weakest motion degree is adaptively selected as the depth correction source. The motion region of the color video is segmented by motion detection. Then, using the accurate time domain consistent information of the color video, the error variation depth value of the static region in the depth video is corrected in time domain consistency. Finally, the time-domain weighted filtering function with low computational complexity is applied to further optimize the corrected depth video, and the depth video with time-domain consistency optimization is obtained. Compared with the original estimated depth video, this algorithm saves the coding rate of 17.48 and 31.75, and improves the subjective quality of the virtual view drawn by the depth map.
【作者单位】： 宁波大学信息科学与工程学院;
【基金】：国家科技支撑计划(2012BAH67F01) 国家自然科学基金重点项目(U1301257)
【分类号】：TP391.41
本文编号：2328752
[Abstract]:The inconsistency of depth time domain exists in the static region of depth video, which leads to the decrease of coding efficiency and affects the quality of rendering. To solve this problem, a depth video consistency enhancement algorithm based on scene motion degree is proposed. Firstly, a block-based histogram difference (Block Histogram Difference,BH) is used to measure the relative motion degree of the depth video between frames. According to the BH value, the video segment with the weakest motion degree is adaptively selected as the depth correction source. The motion region of the color video is segmented by motion detection. Then, using the accurate time domain consistent information of the color video, the error variation depth value of the static region in the depth video is corrected in time domain consistency. Finally, the time-domain weighted filtering function with low computational complexity is applied to further optimize the corrected depth video, and the depth video with time-domain consistency optimization is obtained. Compared with the original estimated depth video, this algorithm saves the coding rate of 17.48 and 31.75, and improves the subjective quality of the virtual view drawn by the depth map.
【作者单位】： 宁波大学信息科学与工程学院;
【基金】：国家科技支撑计划(2012BAH67F01) 国家自然科学基金重点项目(U1301257)
【分类号】：TP391.41
【相似文献】
相关期刊论文 前1条
1 朱修传;;纠正MasterCAM 9帮助中关于拉伸到点的错误[J];机械工程师;2008年01期
，本文编号：2328752
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2328752.html
