摘要
为解决立体图像的差错问题,提出了一种基于边界平滑性准则掩盖方法。首先根据视差相关性为丢失块选择参考块,然后结合像素间色彩空间相似程度和几何空间距离接近程度为它们进行自适应权重的视差匹配,接着给出了边界平滑性的代价函数,并据此度量选择最平滑边界的块作为初步的掩盖效果,最后再次利用边界平滑性准则判断丢失块的属性,并根据其属性采用不同算法对之前的初步掩盖内容进行调整。实验结果表明,相比其他算法,该算法在主、客观质量上均具有较好的优势。
An error concealment algorithm based on boundary smoothness criteria is proposed in order to solve the problem of errors in stereoscopic images.Considering the factors of color similarity and geometric distance proximity,we conduct the adaptive-weight disparity matching for the reference blocks of the lost blocks according to the disparity relativity.We give the cost function of boundary smoothness and select the block with the most smooth boundary as the preliminary concealment result.Finally,we carry out different algorithms to refine the preliminary concealment result according to the property of the lost block which is judged by the boundary smoothness criteria again.Experiment results show that the proposed method can outperform other algorithms in terms of the subjective and objective quality.
出处
《南京邮电大学学报（自然科学版）》
北大核心
2013年第3期45-49,共5页
Journal of Nanjing University of Posts and Telecommunications：Natural Science Edition
基金
国家自然科学基金(61071091)
国家自然科学基金青年科学基金(61001152)
江苏省普通高校研究生科研创新计划(CX10B_188Z)
南京邮电大学科研基金(NY210069)资助项目
关键词
立体图像
边界平滑性
误码掩盖
stereoscopic images
boundary smoothness
error concealment
