一种具有边缘保持的多尺度马尔可夫随机场模型图像分割方法
|
摘要
针对图像分割中常规四叉树结构的多尺度马尔可夫随机场模型非重叠区域在最优化过程中所造成的块效应,以及建模和推理过程导致低分辨率图像边缘细节模糊、缺失的现象,提出了一种具有边缘保持的多尺度马尔可夫随机场(Edge Preserving Multi-Resolution Markov Random Field, EPMRMRF)模型。该模型首先利用邻接区域之间的交互重叠约束,将局部区域的优化传递到相邻区域;其次采用具有边缘保持作用的Cauchy分布提取图像的多尺度边缘先验知识,在不同尺度上实现图像局部区域特征和多尺度边缘特征的融合。之后,为了对EPMRMRF模型进行迭代优化,提出一种分层区域置信度传播算法(Hierarchical Regional Belief Propagation Algorithm,HRBP),基于最大后验准则,求解马尔可夫随机场最大后验全局分布。实验结果表明,EPMRMRF模型和HRBP分割算法不仅有效保持了图像分割结果的边缘,获得了更好的分割结果,而且具有较快的分割速度,概率兰德指数相似性评价指标平均提升至0.890 9,全局一致性误差差异性评价指标平均降低至0.192 3。
For conventional quad tree structure of multi-resolution Markov random field using in image segmentation, blocking effect is caused by optimization process in the overlap regions. Modeling and reasoning processes lead a low resolution image edge with fuzzy or lacking details. In this paper, an image segmentation model of edge-preserving multi-resolution Markov random field(EPMRMRF) is proposed. This model transfers the optimization of the local region into the neighboring region using interact overlap constraint between the adjacent regions. For the fuse of local regional character and multi-scale edge character in different scale images, the Cauchy distribution is utilized to extract image multi-scale edge prior knowledge. And then, a hierarchical regional belief propagation algorithm(HRBP) is proposed to iteratively optimize the EPMRMRF segmentation model by solving the Markov random field maximum a posteriori global distribution based on the maximum a posteriori. Compared with the conventional region-based Markov random field model, experiments show that the proposed algorithm can provide a better segmentation result and faster segmentation rate, in addition, the probabilistic rand index similarity evaluation index rises up to 0.890 9 and the global consistency error difference evaluation index drops to 0.192 3 on the average.
For conventional quad tree structure of multi-resolution Markov random field using in image segmentation, blocking effect is caused by optimization process in the overlap regions. Modeling and reasoning processes lead a low resolution image edge with fuzzy or lacking details. In this paper, an image segmentation model of edge-preserving multi-resolution Markov random field(EPMRMRF) is proposed. This model transfers the optimization of the local region into the neighboring region using interact overlap constraint between the adjacent regions. For the fuse of local regional character and multi-scale edge character in different scale images, the Cauchy distribution is utilized to extract image multi-scale edge prior knowledge. And then, a hierarchical regional belief propagation algorithm(HRBP) is proposed to iteratively optimize the EPMRMRF segmentation model by solving the Markov random field maximum a posteriori global distribution based on the maximum a posteriori. Compared with the conventional region-based Markov random field model, experiments show that the proposed algorithm can provide a better segmentation result and faster segmentation rate, in addition, the probabilistic rand index similarity evaluation index rises up to 0.890 9 and the global consistency error difference evaluation index drops to 0.192 3 on the average.
引文
[1]汪西莉,焦李成.基于多尺度马尔可夫随机场的图像分割[J].计算机科学,2003,30(7):174-176.WANG Xili,JIAO Licheng.Image segmentation based on multiscale Markov random field[J].Computer Science,2003,30(7):174-176.
[2]姚婷婷,谢昭.多层次MRF重标记及映射法则下的图像分割[J].自动化学报,2012,38(9):1-13.YAO Tingting,XIE Zhao.Top-down inference with relabeling and mapping rules in hierarchical MRF for image segmentation[J].Acta Automatica Sinica,2012,38(9):1-13.
[3]ZHENG C,WANG L,CHEN R,et al.Image segmentation using multiregion-resolution MRF model[J].IEEE Geoscience and Remote Sensing Letters,2013,10(4):816-820.
[4]程诗尧,梅天灿,刘国英.顾及结构特征的多层次马尔科夫随机场模型在影像分类中的应用[J].武汉大学学报(信息科学版),2015,40(9):1180-1187.CHENG Shiyao,MEI Tiancan,LIU Guoying.Application of multi-level MRF using structural feature to remote sensing image classification[J].Geomatics and Information Science of Wuhan University,2015,40(9):1180-1187.
[5]FELZENSZWALB P F,HUTTENLOCHER D P.Efficient belief propagation for early vision[J].International Journal of Computer Vision,2006,70(1):167-181.
[6]钱生,陈宗海,林名强,等.基于条件随机场和图像分割的显著性检测[J].自动化学报,2015,41(4):711-724.QIAN Sheng,CHEN Zonghai,LIN Mingqiang,et al.Saliency detection based on conditional random field and image segmentation[J].Acta Automatica Sinica,2015,41(4):711-724.
[7]徐胜军,韩九强,何波,等.融合边缘特征的马尔可夫随机场模型及分割算法[J].西安交通大学学报,2014,48(2):14-19.XU Shengjun,HAN Jiuqiang,HE Bo,et al.A region Markov random field model with integrated edge feature and image segmentation algorithm[J].Journal of Xi’an Jiaotong University,2014,48(2):14-19.
[8]KOHLI P,OSOKIN A,JEGELKA S.A principled deep random field model for image segmentation[C]∥2013IEEE Conference on Computer Vision and Pattern Recognition(CVPR).Piscataway,NJ,USA:IEEE,2013:1971-1978.
[9]张姝茵,侯彪.高概率选择和自适应MRF的极化SAR分类[J].西安电子科技大学学报(自然科学版),2017,44(6):59-64.ZHANG Shuyin,HOU Biao.POLSAR image classification via high-probability selection and adaptive MRF[J].Journal of Xidian University(Natural Science Edition),2017,44(6):59-64.
[10]邓燕子,卢朝阳,李静,等.采用多层图模型推理的道路场景分割算法[J].西安交通大学学报,2017,51(12):62-67.DENG Yanzi,LU Zhaoyang,LI Jing,et al.A segmentation algorithm for road scenes using hierarchical graph based inference[J].Journal of Xi’an Jiaotong University,2017,51(12):62-67.
[11]WANG X,ZHANG W,JI Q.A kernel PCA shape prior and edge based MRF image segmentation[J].Chinese Journal of Electronics,2016,25(5):892-900.
[12]WANG Q.HMRF-EM-image:implementation of the hidden Markov random field model and its expectationmaximization algorithm[J].Computer Science,2012,94(1):222-233.
[13]徐胜军,韩九强,赵亮,等.用于图像分割的局部区域能量最小化算法[J].西安交通大学学报,2011,45(8):7-12.XU Shengjun,HAN Jiuqiang,ZHAO Liang,et al.Algorithm of minimizing local region energy for image segmentation[J].Journal of Xi’an Jiaotong University,2011,45(8):7-12.
[14]徐胜军,韩九强,刘光辉,等.基于局部空间自适应MRF模型的图像分割[J].控制与决策,2013,28(6):889-893.XU Shengjun,HAN Jiuqiang,LIU Guanghui,et al.Image segmentation based on local spatial adaptive Markov field model[J].Control and Decision,2013,28(6):889-893.
[15]LI S Z.Markov random field modeling in computer vision[M].New York,NY,USA:Springer-Verlag,2001:13-16.
[16]YU Q,CLAUSI D A.IRGS:image segmentation using edge penalties and region growing[J].IEEETransactions on Pattern Analysis and Machine Intelligence,2008,30(12):2126-2139.
[17]KAYABOL K,KURUOGLU E,SANKUR B.Bayesian separation of images modeled with MRFs using MCMC[J].IEEE Transactions on Image Processing,2009,18(5):982-994.
[18]YANG Qingxiong,WANG Liang.Stereo matching with color-weighted correlation,hierarchical belief propagation,and occlusion handling[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2009,31(3):1-12.
[19]MIGNOTTE M.A label field fusion Bayesian model and its penalized maximum rand estimator for image segmentation[M].Piscataway,NJ,USA:IEEEPress,2010:1610-1624.
[20]MARTIN D,FOWLKES C,TAL D,et al.A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics[C]∥Proceedings of 8th IEEE International Conference on Computer Vision(ICCV).Piscataway,NJ,USA:IEEE,2002:416-423.
[2]姚婷婷,谢昭.多层次MRF重标记及映射法则下的图像分割[J].自动化学报,2012,38(9):1-13.YAO Tingting,XIE Zhao.Top-down inference with relabeling and mapping rules in hierarchical MRF for image segmentation[J].Acta Automatica Sinica,2012,38(9):1-13.
[3]ZHENG C,WANG L,CHEN R,et al.Image segmentation using multiregion-resolution MRF model[J].IEEE Geoscience and Remote Sensing Letters,2013,10(4):816-820.
[4]程诗尧,梅天灿,刘国英.顾及结构特征的多层次马尔科夫随机场模型在影像分类中的应用[J].武汉大学学报(信息科学版),2015,40(9):1180-1187.CHENG Shiyao,MEI Tiancan,LIU Guoying.Application of multi-level MRF using structural feature to remote sensing image classification[J].Geomatics and Information Science of Wuhan University,2015,40(9):1180-1187.
[5]FELZENSZWALB P F,HUTTENLOCHER D P.Efficient belief propagation for early vision[J].International Journal of Computer Vision,2006,70(1):167-181.
[6]钱生,陈宗海,林名强,等.基于条件随机场和图像分割的显著性检测[J].自动化学报,2015,41(4):711-724.QIAN Sheng,CHEN Zonghai,LIN Mingqiang,et al.Saliency detection based on conditional random field and image segmentation[J].Acta Automatica Sinica,2015,41(4):711-724.
[7]徐胜军,韩九强,何波,等.融合边缘特征的马尔可夫随机场模型及分割算法[J].西安交通大学学报,2014,48(2):14-19.XU Shengjun,HAN Jiuqiang,HE Bo,et al.A region Markov random field model with integrated edge feature and image segmentation algorithm[J].Journal of Xi’an Jiaotong University,2014,48(2):14-19.
[8]KOHLI P,OSOKIN A,JEGELKA S.A principled deep random field model for image segmentation[C]∥2013IEEE Conference on Computer Vision and Pattern Recognition(CVPR).Piscataway,NJ,USA:IEEE,2013:1971-1978.
[9]张姝茵,侯彪.高概率选择和自适应MRF的极化SAR分类[J].西安电子科技大学学报(自然科学版),2017,44(6):59-64.ZHANG Shuyin,HOU Biao.POLSAR image classification via high-probability selection and adaptive MRF[J].Journal of Xidian University(Natural Science Edition),2017,44(6):59-64.
[10]邓燕子,卢朝阳,李静,等.采用多层图模型推理的道路场景分割算法[J].西安交通大学学报,2017,51(12):62-67.DENG Yanzi,LU Zhaoyang,LI Jing,et al.A segmentation algorithm for road scenes using hierarchical graph based inference[J].Journal of Xi’an Jiaotong University,2017,51(12):62-67.
[11]WANG X,ZHANG W,JI Q.A kernel PCA shape prior and edge based MRF image segmentation[J].Chinese Journal of Electronics,2016,25(5):892-900.
[12]WANG Q.HMRF-EM-image:implementation of the hidden Markov random field model and its expectationmaximization algorithm[J].Computer Science,2012,94(1):222-233.
[13]徐胜军,韩九强,赵亮,等.用于图像分割的局部区域能量最小化算法[J].西安交通大学学报,2011,45(8):7-12.XU Shengjun,HAN Jiuqiang,ZHAO Liang,et al.Algorithm of minimizing local region energy for image segmentation[J].Journal of Xi’an Jiaotong University,2011,45(8):7-12.
[14]徐胜军,韩九强,刘光辉,等.基于局部空间自适应MRF模型的图像分割[J].控制与决策,2013,28(6):889-893.XU Shengjun,HAN Jiuqiang,LIU Guanghui,et al.Image segmentation based on local spatial adaptive Markov field model[J].Control and Decision,2013,28(6):889-893.
[15]LI S Z.Markov random field modeling in computer vision[M].New York,NY,USA:Springer-Verlag,2001:13-16.
[16]YU Q,CLAUSI D A.IRGS:image segmentation using edge penalties and region growing[J].IEEETransactions on Pattern Analysis and Machine Intelligence,2008,30(12):2126-2139.
[17]KAYABOL K,KURUOGLU E,SANKUR B.Bayesian separation of images modeled with MRFs using MCMC[J].IEEE Transactions on Image Processing,2009,18(5):982-994.
[18]YANG Qingxiong,WANG Liang.Stereo matching with color-weighted correlation,hierarchical belief propagation,and occlusion handling[J].IEEE Transactions on Pattern Analysis and Machine Intelligence,2009,31(3):1-12.
[19]MIGNOTTE M.A label field fusion Bayesian model and its penalized maximum rand estimator for image segmentation[M].Piscataway,NJ,USA:IEEEPress,2010:1610-1624.
[20]MARTIN D,FOWLKES C,TAL D,et al.A database of human segmented natural images and its application to evaluating segmentation algorithms and measuring ecological statistics[C]∥Proceedings of 8th IEEE International Conference on Computer Vision(ICCV).Piscataway,NJ,USA:IEEE,2002:416-423.