基于Gloabl-Local评估方法的U-Net图像分割
|
摘要
精确的图像分割是完成图像中物体姿态、大小估计的重要步骤,但由于物体的多样性、复杂性等原因,使得图像分割在计算机视觉领域仍然是具有挑战性的任务。针对标准U-Net模型实现端到端的图像分割时精确度不高、训练难以收敛等问题设计了一种基于Gloabl-Local评估方法的U-Net图像分割方法。首先根据同时兼顾全局信息和局部信息能够得到精确的图像分割图,论文提出了Gloabl-Local训练过程易拟合等问题,提出了改进U-Net法和4个公共显著性检测数据集训练改进U-Net网络模型,大大提高了图像分割的准确率。论文的方法平均准确率达到90.74%,与标准U-Net相比具有更好的分割效果。此方法实现了准确高效的图像分割,为估计图像中物体的姿态、大小提供了可靠依据。
Accurate image segmentation is an important step to complete the estimation of object pose and size in images. However,due to the diversity and complexity of objects,image segmentation is still a challenging task in the field of computer vision.According to the inaccuracy of the standard U-Net model in image segmentation and the difficulty of convergence in training,this paper designs a U-Net image segmentation method based on Gloabl-Local evaluation method. First,accurate image segmentation can be obtained according to both global information and local information. This paper proposes a Gloabl-Local evaluation method,which combines global information and local information. Secondly,aiming at the problem of easy fitting of standard U-Net training process,an improved U-Net network model is proposed,which solves the problem of over-fitting in the training process. Gloabl-Local evaluation method and four common saliency detection data sets are used to train the improved U-Net network model. The results show that this method greatly improves the accuracy of image segmentation. The average accuracy of the proposed method reaches 90.74%. This method has higher segmentation accuracy than standard U-Net. This method achieves accurate and efficient image segmentation and provides a reliable basis for estimating the pose and size of objects in the image.
Accurate image segmentation is an important step to complete the estimation of object pose and size in images. However,due to the diversity and complexity of objects,image segmentation is still a challenging task in the field of computer vision.According to the inaccuracy of the standard U-Net model in image segmentation and the difficulty of convergence in training,this paper designs a U-Net image segmentation method based on Gloabl-Local evaluation method. First,accurate image segmentation can be obtained according to both global information and local information. This paper proposes a Gloabl-Local evaluation method,which combines global information and local information. Secondly,aiming at the problem of easy fitting of standard U-Net training process,an improved U-Net network model is proposed,which solves the problem of over-fitting in the training process. Gloabl-Local evaluation method and four common saliency detection data sets are used to train the improved U-Net network model. The results show that this method greatly improves the accuracy of image segmentation. The average accuracy of the proposed method reaches 90.74%. This method has higher segmentation accuracy than standard U-Net. This method achieves accurate and efficient image segmentation and provides a reliable basis for estimating the pose and size of objects in the image.
引文
[1]Otsu N.A Threshold Selection Method from Gray-Level Histograms[J].IEEE Transactions on Systems,Man,and Cybernetics,1979,9(1):62-66.
[2]Lakshmi S,Sankaranarayanan D V.A study of edge detection techniques for segmentation computing approaches[J].IJCA Special Issue on“Computer Aided Soft Computing Techniques for Imaging and Biomedical Applications”CASCT,2010:35-40.
[3]Adams R,Bischof L.Seeded region growing[J].IEEETransactions on pattern analysis and machine intelligence,1994,16(6):641-647.
[4]Selver M A,Kocao?lu A,Demir G K,et al.Patient oriented and robust automatic liver segmentation for pre-evaluation of liver transplantation[J].Computers in Biology and Medicine,2008,38(7):765-784.
[5]黄琴波.结合特定理论的图像分割方法[J].电子科技,2010,23(12):92-95.HUANG Qinbo.Image segmentation Method Cobined with Specific Theory[J].Electronic Technology,2010,23(12):92-95.
[6]Long J,Shelhamer E,Darrell T.Fully convolutional networks for semantic segmentation[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.2015:3431-3440.
[7]Chen L C,Papandreou G,Kokkinos I,et al.Deeplab:Semantic image segmentation with deep convolutional nets,atrous convolution,and fully connected crfs[J].IEEEtransactions on pattern analysis and machine intelligence,2018,40(4):834-848.
[8]Lin G,Shen C,Reid I,et al.Deeply learning the messages in message passing inference[C]//Advances in Neural Information Processing Systems.2015:361-369.
[9]Shore J,Johnson R.Axiomatic derivation of the principle of maximum entropy and the principle of minimum cross-entropy[J].IEEE Transactions on information theory,1980,26(1):26-37.
[10]Arbelaez P,Maire M,Fowlkes C,et al.Contour detection and hierarchical image segmentation[J].IEEE transactions on pattern analysis and machine intelligence,2011,33(5):898-916.
[11]Jaccard P.étude comparative de la distribution florale dans une portion des Alpes et des Jura[J].Bull Soc Vaudoise Sci Nat,1901,37:547-579.
[12]Benesty J,Chen J,Huang Y,et al.Pearson correlation coefficient[M].Noise reduction in speech processing.Springer,Berlin,Heidelberg,2009:1-4.943
[13]Ronneberger O,Fischer P,Brox T.U-net:Convolutional networks for biomedical image segmentation[C]//International Conference on Medical image computing and computer-assisted intervention.Springer,Cham,2015:234-241.
[14]Wang Z,Bovik A C,Sheikh H R,et al.Image quality assessment:from error visibility to structural similarity[J].IEEE transactions on image processing,2004,13(4):600-612.
[15]Rokach L,Maimon O.Clustering methods[M].Data mining and knowledge discovery handbook.Springer,Boston,MA,2005:321-352.
[16]Han B,Zhu H,Ding Y.Bottom-up saliency based on weighted sparse coding residual[C]//Proceedings of the19th ACM international conference on Multimedia.ACM,2011:1117-1120.
[17]Pfeiffenberger H,Carlson D."Earth System Science Data"(ESSD)-A Peer Reviewed Journal for Publication of Data[J].D-Lib Magazine,2011,17(1/2).
[18]Tong N,Lu H,Ruan X,et al.Salient object detection via bootstrap learning[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.2015:1884-1892.
[19]Li G,Yu Y.Visual saliency based on multiscale deep features[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.2015:5455-5463.
[2]Lakshmi S,Sankaranarayanan D V.A study of edge detection techniques for segmentation computing approaches[J].IJCA Special Issue on“Computer Aided Soft Computing Techniques for Imaging and Biomedical Applications”CASCT,2010:35-40.
[3]Adams R,Bischof L.Seeded region growing[J].IEEETransactions on pattern analysis and machine intelligence,1994,16(6):641-647.
[4]Selver M A,Kocao?lu A,Demir G K,et al.Patient oriented and robust automatic liver segmentation for pre-evaluation of liver transplantation[J].Computers in Biology and Medicine,2008,38(7):765-784.
[5]黄琴波.结合特定理论的图像分割方法[J].电子科技,2010,23(12):92-95.HUANG Qinbo.Image segmentation Method Cobined with Specific Theory[J].Electronic Technology,2010,23(12):92-95.
[6]Long J,Shelhamer E,Darrell T.Fully convolutional networks for semantic segmentation[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.2015:3431-3440.
[7]Chen L C,Papandreou G,Kokkinos I,et al.Deeplab:Semantic image segmentation with deep convolutional nets,atrous convolution,and fully connected crfs[J].IEEEtransactions on pattern analysis and machine intelligence,2018,40(4):834-848.
[8]Lin G,Shen C,Reid I,et al.Deeply learning the messages in message passing inference[C]//Advances in Neural Information Processing Systems.2015:361-369.
[9]Shore J,Johnson R.Axiomatic derivation of the principle of maximum entropy and the principle of minimum cross-entropy[J].IEEE Transactions on information theory,1980,26(1):26-37.
[10]Arbelaez P,Maire M,Fowlkes C,et al.Contour detection and hierarchical image segmentation[J].IEEE transactions on pattern analysis and machine intelligence,2011,33(5):898-916.
[11]Jaccard P.étude comparative de la distribution florale dans une portion des Alpes et des Jura[J].Bull Soc Vaudoise Sci Nat,1901,37:547-579.
[12]Benesty J,Chen J,Huang Y,et al.Pearson correlation coefficient[M].Noise reduction in speech processing.Springer,Berlin,Heidelberg,2009:1-4.943
[13]Ronneberger O,Fischer P,Brox T.U-net:Convolutional networks for biomedical image segmentation[C]//International Conference on Medical image computing and computer-assisted intervention.Springer,Cham,2015:234-241.
[14]Wang Z,Bovik A C,Sheikh H R,et al.Image quality assessment:from error visibility to structural similarity[J].IEEE transactions on image processing,2004,13(4):600-612.
[15]Rokach L,Maimon O.Clustering methods[M].Data mining and knowledge discovery handbook.Springer,Boston,MA,2005:321-352.
[16]Han B,Zhu H,Ding Y.Bottom-up saliency based on weighted sparse coding residual[C]//Proceedings of the19th ACM international conference on Multimedia.ACM,2011:1117-1120.
[17]Pfeiffenberger H,Carlson D."Earth System Science Data"(ESSD)-A Peer Reviewed Journal for Publication of Data[J].D-Lib Magazine,2011,17(1/2).
[18]Tong N,Lu H,Ruan X,et al.Salient object detection via bootstrap learning[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.2015:1884-1892.
[19]Li G,Yu Y.Visual saliency based on multiscale deep features[C]//Proceedings of the IEEE conference on computer vision and pattern recognition.2015:5455-5463.