基于One-class SVM的噪声图像分割方法
|
摘要
为解决现有无监督图像分割模型对强噪声环境鲁棒性差、无法适应复杂混合噪声的问题,提出了一种基于One-class SVM方法的改进后的噪声鲁棒图像分割模型。首先,基于One-class SVM构建一种数据离群程度检测机制;然后,将离群程度值引入能量泛函,令分割模型可以在多种噪声强度下获得较为准确的图像信息,同时避免现有方法在强噪声环境下,降权机制失效的问题;最后,通过最小化能量函数,驱动分割轮廓向目标边缘演化。在噪声图像分割实验中,当选取不同类型和强度的噪声时,该模型均能得到较为理想的分割结果。在F_1-score评估标准下,该模型比基于局部相关熵的K-means(LCK)模型高0.2～0.3,在强噪声环境下具有更高的稳定性,且在分割收敛时间上仅略大于LCK模型0.1 s左右。实验结果表明,所提模型在未显著增加分割耗时的前提下,对于概率、极值及混合噪声均有着更强的鲁棒性,并且可以分割带有噪声的自然图像。
To deal with poor robustness in strong noise environment, weak adaptability to complex mixed noise that appear in the existing unsupervised image segmentation models, an improved noise-robust image segmentation model based on One-class SVM(Support Vector Machine) method was proposed. Firstly, a data outlier detection mechanism was constructed based on One-class SVM. Secondly, an outlier degree was introduced into the energy function, so that more accurate image information could be obtained by the proposed model under multiple noise intensities and the failure of weight-descend mechanism in strong noise environment was avoided. Finally, the segmentation contour was driven to the target edge by minimizing the energy function. In noise image segmentation experiments, the proposed model could obtain ideal segmentation results with different types and intensities of noise. Under F_1-score metric, the proposed model is 0.2 to 0.3 higher than LCK(Local Correntropy-based K-means) model, and has better stability in strong noise environments. The segmentation convergence time of the proposed model is only slightly longer than that of LCK model by about 0.1 s. Experimental results show that the proposed model is more robust to probabilistic, extreme values and mixed noise without significantly increase of segmentation time, and can segment natural images with noise.
To deal with poor robustness in strong noise environment, weak adaptability to complex mixed noise that appear in the existing unsupervised image segmentation models, an improved noise-robust image segmentation model based on One-class SVM(Support Vector Machine) method was proposed. Firstly, a data outlier detection mechanism was constructed based on One-class SVM. Secondly, an outlier degree was introduced into the energy function, so that more accurate image information could be obtained by the proposed model under multiple noise intensities and the failure of weight-descend mechanism in strong noise environment was avoided. Finally, the segmentation contour was driven to the target edge by minimizing the energy function. In noise image segmentation experiments, the proposed model could obtain ideal segmentation results with different types and intensities of noise. Under F_1-score metric, the proposed model is 0.2 to 0.3 higher than LCK(Local Correntropy-based K-means) model, and has better stability in strong noise environments. The segmentation convergence time of the proposed model is only slightly longer than that of LCK model by about 0.1 s. Experimental results show that the proposed model is more robust to probabilistic, extreme values and mixed noise without significantly increase of segmentation time, and can segment natural images with noise.
引文
[1]KASS M,WITKIN A,TERZOPOULOS D.Snakes:active contour models[J].International Journal of Computer Vision,1988,1(4):321-331.
[2]CHAN T F,VESE L A.Active contour without edges[J].IEEETransactions on Image Processing,2001,10(2):266-277.
[3]LI C,KAO C-Y,GORE J C,et al.Implicit active contours driven by local binary fitting energy[C]//CVPR 2007:Proceedings of the2007 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.Washington,DC:IEEE Computer Society,2007:1-7
[4]ZHANG K,SONG H,ZHANG L.Active contours driven by local image fitting energy[J].Pattern Recognition,2010,43(4):1199-1206.
[5]WANG L,PAN C.Robust level set image segmentation via a local correntropy-based K-means clustering[J].Pattern Recognition,2014,47(5):1917-1925.
[6]JIANG X-L,WANG Q,HE B,et al.Robust level set image segmentation algorithm using local correntropy-based fuzzy c-means clustering with spatial constraints[J].Neurocomputing,2016,207(C):22-35.
[7]LI C,XU C,GUI C,et al.Level set evolution without re-initialization:a new variational formulation[C]//CVPR'05:Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.Washington,DC:IEEE Computer Society,2005:430-436.
[8]赵怡,邓红霞,张玲,等.基于最大类间方差的权重自适应活动轮廓模型[J].计算机工程与设计,2018,39(2):486-491.(ZHAOY,DENG H X,ZHANG L,et al.Weight-self adjustment active contour model based on method of maximum classes square error[J].Computer Engineering and Design,2018,39(2):486-491.)
[9]李钢,李海芳,尚方信,等.基于梯度信息的自适应邻域噪声图像分割模型[J].计算机工程,2018,44(5):227-233.(LI G,LI HF,SHANG F X,et al.Noise image segmentation model with adaptive neighborhood based on gradient information[J].Computer Engineering,2018,44(5):227-233.)
[10]SCHLKOPF B,PLATT J C,SHAWE-TAYLOR J,et al.Estimating the support of a high-dimensional distribution[J].Neural Computation,2001,13(7):1443-1471.
[11]李昊奇,应娜,郭春生,等.基于深度信念网络和线性单分类SVM的高维异常检测[J].电信科学,2018,34(1):34-42.(LIH Q,YING N,GUO C S,et al.High-dimensional outlier detection based on deep belief network and linear one-class SVM[J].Telecommunications Science,2018,34(1):34-42.)
[12]杨成佳.图像去噪及其效果评估若干问题研究[D].长春:吉林大学,2016:4-5.(YANG C J.Research on image denoising and its effect evaluation[D].Changchun:Jilin University,2016:4-5.)
[13]Le CUN Y,BOTTOU L,BENGIO Y,et al.MNIST[DB/OL].(2012-01-01)[2018-08-27].http://yann.lecun.com/exdb/mnist/.
[14]BORENSTEIN E.Weizmann horse database[DB/OL].(2005-01-19)[2018-08-27].http://www.msri.org/m/people/members/eranb/.
[2]CHAN T F,VESE L A.Active contour without edges[J].IEEETransactions on Image Processing,2001,10(2):266-277.
[3]LI C,KAO C-Y,GORE J C,et al.Implicit active contours driven by local binary fitting energy[C]//CVPR 2007:Proceedings of the2007 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.Washington,DC:IEEE Computer Society,2007:1-7
[4]ZHANG K,SONG H,ZHANG L.Active contours driven by local image fitting energy[J].Pattern Recognition,2010,43(4):1199-1206.
[5]WANG L,PAN C.Robust level set image segmentation via a local correntropy-based K-means clustering[J].Pattern Recognition,2014,47(5):1917-1925.
[6]JIANG X-L,WANG Q,HE B,et al.Robust level set image segmentation algorithm using local correntropy-based fuzzy c-means clustering with spatial constraints[J].Neurocomputing,2016,207(C):22-35.
[7]LI C,XU C,GUI C,et al.Level set evolution without re-initialization:a new variational formulation[C]//CVPR'05:Proceedings of the 2005 IEEE Computer Society Conference on Computer Vision and Pattern Recognition.Washington,DC:IEEE Computer Society,2005:430-436.
[8]赵怡,邓红霞,张玲,等.基于最大类间方差的权重自适应活动轮廓模型[J].计算机工程与设计,2018,39(2):486-491.(ZHAOY,DENG H X,ZHANG L,et al.Weight-self adjustment active contour model based on method of maximum classes square error[J].Computer Engineering and Design,2018,39(2):486-491.)
[9]李钢,李海芳,尚方信,等.基于梯度信息的自适应邻域噪声图像分割模型[J].计算机工程,2018,44(5):227-233.(LI G,LI HF,SHANG F X,et al.Noise image segmentation model with adaptive neighborhood based on gradient information[J].Computer Engineering,2018,44(5):227-233.)
[10]SCHLKOPF B,PLATT J C,SHAWE-TAYLOR J,et al.Estimating the support of a high-dimensional distribution[J].Neural Computation,2001,13(7):1443-1471.
[11]李昊奇,应娜,郭春生,等.基于深度信念网络和线性单分类SVM的高维异常检测[J].电信科学,2018,34(1):34-42.(LIH Q,YING N,GUO C S,et al.High-dimensional outlier detection based on deep belief network and linear one-class SVM[J].Telecommunications Science,2018,34(1):34-42.)
[12]杨成佳.图像去噪及其效果评估若干问题研究[D].长春:吉林大学,2016:4-5.(YANG C J.Research on image denoising and its effect evaluation[D].Changchun:Jilin University,2016:4-5.)
[13]Le CUN Y,BOTTOU L,BENGIO Y,et al.MNIST[DB/OL].(2012-01-01)[2018-08-27].http://yann.lecun.com/exdb/mnist/.
[14]BORENSTEIN E.Weizmann horse database[DB/OL].(2005-01-19)[2018-08-27].http://www.msri.org/m/people/members/eranb/.