摘要
为了改进视频压缩感知方案的性能,提出了一种基于多重假设的视频压缩感知分层重建方案。该重建方案以图像组为单位进行,首先对关键帧进行空域多重假设预测和残差重建;接下来对图像组中的每个非关键帧分配重建层级,并按照由低至高的顺序逐层重建;每个非关键帧的重建过程逐块进行,需要其参考帧中的时域假设预测集合及当前帧中的空域假设预测集合来对每个图像块做混合多重假设预测,并通过求解全变分最小化问题进行残差重建。实验结果表明,在相同采样率下,与已有的视频压缩感知方案相比,本文方案可以获得质量更高的重建图像。
To improve the performance of compressed video sensing （CVS）, a multi-hypothe- sis based hierarchical reconstruction method is proposed. In the presented framework, the key frame in group of picture （GOP） is first predicted by the spatial multi-hypothesis （MH） meth- od and then reconstructed by the residual reconstruction method. Afterwards, the reconstruc- tion levels are allocated to all the non-key frames in current GOP, and then the reconstruction is processed from the lowest level to the highest one. When reconstructing a non-key frame, block by block reconstruction is carried out. For a target block, both of the temporal data set in reference frames and spatial date set in current frame are used to form a MH prediction. Af- ter that, the total variation minimization problem is solved to execute residual reconstruction. Experimental results show that compared with the existing CVS methods, the proposed one can obtain higher quality of reconstructed images at the same sampling rates.
出处
《数据采集与处理》
CSCD
北大核心
2013年第6期730-738,共9页
Journal of Data Acquisition and Processing
基金
国家自然科学基金(61261023)资助项目
广西自然科学基金(2011GXNSFD018024
2013GXNSFBA019272)资助项目
广西教育厅科研项目(201203YB001)资助项目
关键词
压缩感知
多重假设
分层重建
全变分最小化
compressed sensing
multi-hypothesis
hierarchical reconstruction
total variation minimization
