基于加权非局部相似性的视频压缩感知多假设重构算法
|
摘要
分布式视频压缩感知(Distributed Compressed Video Sensing,DCVS)多假设重构算法将传统视频编码中的多假设预测运动估计思想引入到分布式压缩感知视频编码系统中,改善了对视频序列的重构质量。在该算法中,大变化块采用本帧邻域块信息作为参考,而当本帧邻域块含有较多纹理和细节时,算法性能有待提高。为此,对非局部相似性的思想进行改进,提出基于加权非局部相似性的分布式视频压缩感知多假设重构算法。在该算法中,对大变化块中的纹理块采用加权非局部相似性在相邻已重构帧中寻找自相似块,最终生成辅助重构信息块;对于非纹理块,则简单利用加权非局部相似性生成相似块。对不同特点的视频序列的仿真实验结果表明,改进后的算法有效改善了视频序列的重构质量,具有较优的重构SSIM,PSNR指标,其中PSNR约提高1dB。
Multi-hypothesis reconstruction algorithm of DCVS(Distributed Compressed Video Sensing)introduces the idea of multi-hypothesis prediction motion estimation of traditional video encoding into the DCVS encoding system,thus improving the reconstruction quality for video sequence.In this algorithm,the blocks with big changes adopt the information of current frame neighborhood blocks as a reference,and its performance needs to be improved when the neighborhood of frame contains lots of textures and details.Through improving the idea of non-local similarity,this paper proposed a multi-hypothesis reconstruction algorithm of DCVS based on weighted non-local similarity.In the improved algorithm,the weighted non-local similarity is adopted to search the self-similar blocks in adjacent reconstructed frames for the texture block in the block with big changes,finally generating supplementary reconstruction information blocks.For text non-texture blocks,the weighted non-local similarity is utilized to generate similar blocks.For the blocks with small changes,inter-frame multi-hypothesis reconstruction is adopted,and non-critical frame reconstruction is assisted.Simulation results based on different video sequences show that the proposed algorithm can improve the reconstruction quality of video sequence effectively,and has better reconstructed SSIM and PSNR,and the PSNR is about 1dB higher.
Multi-hypothesis reconstruction algorithm of DCVS(Distributed Compressed Video Sensing)introduces the idea of multi-hypothesis prediction motion estimation of traditional video encoding into the DCVS encoding system,thus improving the reconstruction quality for video sequence.In this algorithm,the blocks with big changes adopt the information of current frame neighborhood blocks as a reference,and its performance needs to be improved when the neighborhood of frame contains lots of textures and details.Through improving the idea of non-local similarity,this paper proposed a multi-hypothesis reconstruction algorithm of DCVS based on weighted non-local similarity.In the improved algorithm,the weighted non-local similarity is adopted to search the self-similar blocks in adjacent reconstructed frames for the texture block in the block with big changes,finally generating supplementary reconstruction information blocks.For text non-texture blocks,the weighted non-local similarity is utilized to generate similar blocks.For the blocks with small changes,inter-frame multi-hypothesis reconstruction is adopted,and non-critical frame reconstruction is assisted.Simulation results based on different video sequences show that the proposed algorithm can improve the reconstruction quality of video sequence effectively,and has better reconstructed SSIM and PSNR,and the PSNR is about 1dB higher.
引文
[1]SHEN Y C,CHENG H P,LUO Y H.Efficient Real-Time Distributed Video Coding by Parallel Progressive Side Information Regeneration[J].IEEE Sensors Journal,2017,17(6):1872-1883.
[2]LI W K,LU C S.Distributed Video Compressive Sensing[C]∥IEEE International Conference on Acoustics.2009:1169-1172.
[3]CORALIA C,ANDREW T.Quantitative Recovery Conditions for Tree-Based Compressed Sensing[J].IEEE Transactions on Information Theory,2017,63(3):1555-1571.
[4]XU J,ZHANG Y,FU Z Z,et al.Perceptual Distributed Compressive Video Sensing via Reweighted Sampling and Rate-Distortion Optimized Measurements Allocation[J].IEICE Transactions on Information&Systems,2017,100(4):918-922.
[5]SAMAD R,JAFAR Z,SHOTORBAN B B.Distributed Compressed Video Sensing Based on Recursive Least Square Dictionary Learning[C]∥Iranian Conference on Electrical Engineering.2016:1775-1779.
[6]DENG S Y,WANG A H.Residual Distributed Compressive Video Sensing[J].Application Research of Computers,2012,29(4):1553-1556.(in Chinese)邓世洋,王安红.残差分布式视频压缩感知[J].计算机应用研究,2012,29(4):1553-1556.
[7]MENG Y,ZHU J X,ZHANG Y.Multihypothesis Prediction Algorithm of DCVS based on Spatio Temporal Correlativity[J].Application Research of Computers,2014,31(2):637-640.(in Chinese)孟雨,朱金秀,张瑶.基于时空相关性的分布式压缩感知多假设预测重构算法[J].计算机应用研究,2014,31(2):637-640.
[8]ZHANG J,DONG Y N.Distributed Video Compressive Sensing Reconstruction Based on Motion aligned Predictive Model[J].Journal of Nanjing University of Posts and Telecommunications,2014,34(4):63-71.(in Chinese)张健,董育宁.基于运动对齐预测模型的分布式视频压缩感知重构[J].南京邮电大学学报,2014,34(4):63-71.
[9]WU Q S,FANG S L.Structured Bayesian Compressive Sensing with Spatial Location Dependence via Variational Bayesian Inference[J].Digital Signal Processing,2017,71(4):95-107.
[10]ZHU J X,ZHANG Y,ZHANG X W.Reconstruction using Interpolation Multihypothesis Prediction for DCVS[J].Computer Engineering and Design,2016,37(2):443-449.(in Chinese)朱金秀,张瑶,张学武.DCVS中插值多假设预测重构算法[J].计算机工程与设计,2016,37(2):443-449.
[11]CHEN W B,GAO X W,FAN X P,et al.Spatial-temporal Recovery for Hierarchical Frame Based Video Compressed Sensing[C]∥IEEE Conference on Image Processing.2015:1110-1114.
[12]CHEN S Z,LI G Y,LIAN Q S.Image Compressed Sensing Based on Nonlocal Similarity and Alternating Iterative Optimization Algorithm[J].Signal Porcessing,2012,28(2):200-205.(in Chinese)陈书贞,李光耀,练秋生.基于非局部相似性和交替迭代优化算法的图像压缩感知[J].信号处理,2012,28(2):200-205.
[13]ZHANGM L,DESROSIERS C.Robust MRI Reconstruction via Re-weighted Total Variation and Non-local Sparse Regression[C]∥IEEE Multimedia Signal Processing.2016:1-6.
[14]REN C,HE X.Single Image Super-resolution using Local Geometric Duality and Non-Local Similarity[J].IEEE Transactions on Image Processing,2016,25(5):2168-2183.
[15]KUO Y H,WU K,CHEN J.A Scheme for Distributed Compressed Video Sensing Basedon Hypothesis Set Optimization Techniques[J].Multidimensional Systems and Signal Processing,2017,28(1):129-148.
[16]WANG L,FENG Y.Compressed Sensing Reconstruction of Hyperspectral Images Based on Spatial-spectral Multi-hypothesis Prediction[J].Journal of Electronics&Information Technology,2015,37(12):3000-3008.(in Chinese)王丽,冯燕.基于空谱联合的多假设预测高光谱图像压缩感知重构算法[J].电子与信息学报,2015,37(12):3000-3008.
[17]CHEN C,ZHANG D Y.Resample-Based Hybrid Multi-Hypothesis Scheme for Distributed Compressive Video Sensing[J].IEICE Transactions on Information&Systems,2017,100(12):3073-3076.
[18]SUNGKWANG M,JAMES E F.Block Compressed Sensing of Images Using Directional Transforms[C]∥Conference on Image Processing.2009:3021-3024.
[19]WANG Z,LIU L H.Research Based on SAR Imaging Technology Sparse Reconstruction[J].Application Research of Computers,2014,21(5):161-166.(in Chinese)王哲,刘力辉.基于稀疏重构的SAR成像技术研究[J].计算机应用,2014,21(5):161-166.
[20]AN W,LIU K,WANG J.Research on Multi-Hypothesis Residual Reconstruction Algorithm Based on Adaptive Sampling[J].Acts Automatica Sinica,2017,43(X):1-11.(in Chinese)安文,刘昆,王杰.基于自适应采样的多假设预测残差重构算法研究[J].自动化学报,2017,43(X):1-11.
[21]XUE T,DONG X D,SHI Y.Multiple Access and Data Reconstruction in Wireless Sensor Networks based on Compressed Sensing[J].IEEE Transactions on Wireless Communication,2013,12(7):3399-3411.
[22]WU M H,LI R,CHEN R.Distributed Video Compressive Sensing Reconstruction Based on Adaptive PCA Sparse Basis[J].Video Engineering,2015,39(2):61-65.(in Chinese)武明虎,李然,陈瑞.自适应PCA稀疏基底的分布式视频压缩感知重构[J].电视技术,2015,39(2):61-65.
[2]LI W K,LU C S.Distributed Video Compressive Sensing[C]∥IEEE International Conference on Acoustics.2009:1169-1172.
[3]CORALIA C,ANDREW T.Quantitative Recovery Conditions for Tree-Based Compressed Sensing[J].IEEE Transactions on Information Theory,2017,63(3):1555-1571.
[4]XU J,ZHANG Y,FU Z Z,et al.Perceptual Distributed Compressive Video Sensing via Reweighted Sampling and Rate-Distortion Optimized Measurements Allocation[J].IEICE Transactions on Information&Systems,2017,100(4):918-922.
[5]SAMAD R,JAFAR Z,SHOTORBAN B B.Distributed Compressed Video Sensing Based on Recursive Least Square Dictionary Learning[C]∥Iranian Conference on Electrical Engineering.2016:1775-1779.
[6]DENG S Y,WANG A H.Residual Distributed Compressive Video Sensing[J].Application Research of Computers,2012,29(4):1553-1556.(in Chinese)邓世洋,王安红.残差分布式视频压缩感知[J].计算机应用研究,2012,29(4):1553-1556.
[7]MENG Y,ZHU J X,ZHANG Y.Multihypothesis Prediction Algorithm of DCVS based on Spatio Temporal Correlativity[J].Application Research of Computers,2014,31(2):637-640.(in Chinese)孟雨,朱金秀,张瑶.基于时空相关性的分布式压缩感知多假设预测重构算法[J].计算机应用研究,2014,31(2):637-640.
[8]ZHANG J,DONG Y N.Distributed Video Compressive Sensing Reconstruction Based on Motion aligned Predictive Model[J].Journal of Nanjing University of Posts and Telecommunications,2014,34(4):63-71.(in Chinese)张健,董育宁.基于运动对齐预测模型的分布式视频压缩感知重构[J].南京邮电大学学报,2014,34(4):63-71.
[9]WU Q S,FANG S L.Structured Bayesian Compressive Sensing with Spatial Location Dependence via Variational Bayesian Inference[J].Digital Signal Processing,2017,71(4):95-107.
[10]ZHU J X,ZHANG Y,ZHANG X W.Reconstruction using Interpolation Multihypothesis Prediction for DCVS[J].Computer Engineering and Design,2016,37(2):443-449.(in Chinese)朱金秀,张瑶,张学武.DCVS中插值多假设预测重构算法[J].计算机工程与设计,2016,37(2):443-449.
[11]CHEN W B,GAO X W,FAN X P,et al.Spatial-temporal Recovery for Hierarchical Frame Based Video Compressed Sensing[C]∥IEEE Conference on Image Processing.2015:1110-1114.
[12]CHEN S Z,LI G Y,LIAN Q S.Image Compressed Sensing Based on Nonlocal Similarity and Alternating Iterative Optimization Algorithm[J].Signal Porcessing,2012,28(2):200-205.(in Chinese)陈书贞,李光耀,练秋生.基于非局部相似性和交替迭代优化算法的图像压缩感知[J].信号处理,2012,28(2):200-205.
[13]ZHANGM L,DESROSIERS C.Robust MRI Reconstruction via Re-weighted Total Variation and Non-local Sparse Regression[C]∥IEEE Multimedia Signal Processing.2016:1-6.
[14]REN C,HE X.Single Image Super-resolution using Local Geometric Duality and Non-Local Similarity[J].IEEE Transactions on Image Processing,2016,25(5):2168-2183.
[15]KUO Y H,WU K,CHEN J.A Scheme for Distributed Compressed Video Sensing Basedon Hypothesis Set Optimization Techniques[J].Multidimensional Systems and Signal Processing,2017,28(1):129-148.
[16]WANG L,FENG Y.Compressed Sensing Reconstruction of Hyperspectral Images Based on Spatial-spectral Multi-hypothesis Prediction[J].Journal of Electronics&Information Technology,2015,37(12):3000-3008.(in Chinese)王丽,冯燕.基于空谱联合的多假设预测高光谱图像压缩感知重构算法[J].电子与信息学报,2015,37(12):3000-3008.
[17]CHEN C,ZHANG D Y.Resample-Based Hybrid Multi-Hypothesis Scheme for Distributed Compressive Video Sensing[J].IEICE Transactions on Information&Systems,2017,100(12):3073-3076.
[18]SUNGKWANG M,JAMES E F.Block Compressed Sensing of Images Using Directional Transforms[C]∥Conference on Image Processing.2009:3021-3024.
[19]WANG Z,LIU L H.Research Based on SAR Imaging Technology Sparse Reconstruction[J].Application Research of Computers,2014,21(5):161-166.(in Chinese)王哲,刘力辉.基于稀疏重构的SAR成像技术研究[J].计算机应用,2014,21(5):161-166.
[20]AN W,LIU K,WANG J.Research on Multi-Hypothesis Residual Reconstruction Algorithm Based on Adaptive Sampling[J].Acts Automatica Sinica,2017,43(X):1-11.(in Chinese)安文,刘昆,王杰.基于自适应采样的多假设预测残差重构算法研究[J].自动化学报,2017,43(X):1-11.
[21]XUE T,DONG X D,SHI Y.Multiple Access and Data Reconstruction in Wireless Sensor Networks based on Compressed Sensing[J].IEEE Transactions on Wireless Communication,2013,12(7):3399-3411.
[22]WU M H,LI R,CHEN R.Distributed Video Compressive Sensing Reconstruction Based on Adaptive PCA Sparse Basis[J].Video Engineering,2015,39(2):61-65.(in Chinese)武明虎,李然,陈瑞.自适应PCA稀疏基底的分布式视频压缩感知重构[J].电视技术,2015,39(2):61-65.