一种离焦模糊半径估计的新方法
|
摘要
离焦模糊半径在离焦模糊图像复原中有着十分重要的作用。根据离焦模糊图像的模糊半径与它的傅里叶频谱图的零点暗圆的半径成反比,文中提出了一种基于支持向量机(SVM)的离焦模糊半径估计的新方法。首先,获取已知模糊半径的模糊图像,将模糊图像频谱图第二个暗圆的半径和已知的模糊半径作为训练样本;然后,基于训练样本,通过支持向量机学习得到模糊半径与第二暗圆半径之间的关系模型;最后,利用关系模型来预测未知的模糊半径。实验表明,该方法简单有效,在图像噪声较小时,估计误差率小于0.4%,优于现有的估计方法。
The blurred radius plays a very important role in the defocused blurred image restoration.According to the inversely proportional relationship between the blurred radius of defocused image and the radius of its Fourier spectrum zero dark circle,a new approach of estimation of radius of defocused blurred image based on support vector machine is proposed. Firstly,it obtains the blurred images whose blurred radus are known and uses the radii of the second spectrum dark circles and the known radus as the training samples. Secondly,based on the training samples,the model of the relationship between the blurred radius and the radius of the second dark circle is obtained by support vector machine. Finally,it uses the model to predict the unknown blurred radius. The experimental results show that this method is simple and effective,and when the noise of the image is very small,the estimation error rate is less than0. 4% which is better than the existing estimation methods.
The blurred radius plays a very important role in the defocused blurred image restoration.According to the inversely proportional relationship between the blurred radius of defocused image and the radius of its Fourier spectrum zero dark circle,a new approach of estimation of radius of defocused blurred image based on support vector machine is proposed. Firstly,it obtains the blurred images whose blurred radus are known and uses the radii of the second spectrum dark circles and the known radus as the training samples. Secondly,based on the training samples,the model of the relationship between the blurred radius and the radius of the second dark circle is obtained by support vector machine. Finally,it uses the model to predict the unknown blurred radius. The experimental results show that this method is simple and effective,and when the noise of the image is very small,the estimation error rate is less than0. 4% which is better than the existing estimation methods.
引文
[1]郑楚君,李榕,常鸿森.离焦模糊数字图像的Wiener滤波频域复原[J].激光杂志,2004,25(5):57-58.
[2]周箩鱼,张葆,杨扬.采用Hough变换的离焦模糊参数的估计[J].红外与激光工程,2012,41(10):2833-2837.
[3]姜明勇,陈向宁,喻夏琼.一种离焦模糊遥感图像盲复原方法[J].测绘科学,2012,37(4):135-137.
[4]梁敏,朱虹.基于边缘模糊频谱特征的散焦参数估计方法[J].计算机应用,2014,34(4):1177-1181.
[5]Jiang Y G,Wu Q,Guo P.Defocused image restoration using RBF network and Kalman filter[C].Proceedings of the 2005 IEEE International Conference on Systems,Man and Cybernetics,2005:2507-2511.
[6]Su L Y,Li F L,Xu F,et al.Defocused image restoration using RBF network and iterative Wiener filter in wavelet domain[C].Proceedings of the first International Congress on Image and Signal Processing.2008:311-315.
[7]Lee J,Fathi A S,Song S.Defocus Blur Estimation Using a Cellular Neural Network[C].Proceedings of the IEEE 12thInternational Workshop on Cellular Nanoscale Networks&Their Applications.2010:1-4.
[8]Moghaddam M E.A mathematical model to estimate out of focus blur[C].Proceedings of the 5th International Symposium on Image and Signal Processing and Analysis,2007:278-281.
[9]Moghaddam M E.A robust noise independent method to estimate out of focus blur[C].Proceedings of the IEEE International Conference on Acoustics,Speech,and Signal Processing.2008:1273-1276.
[10]杨亚威,耿志,王蕊,等.传统相机的空间变化离焦去模糊[J].电光与控制,2015,22(9):91-95.
[11]Gonzalez L,Velasco F,Angulo C,et al.A study on output normalization in multiclass SVMs[J].Pattern Recognition Letters,2013,34(3):344-348.
[12]李航.统计学习方法[M].北京:清华大学出版社,2012:96-135.
[13]Chang C C,Lin C J.LIBSVM:A library for support vector machines[EB/OL].http:∥www.csie.ntu.edu.tw/~cjlin/libsvm.
[2]周箩鱼,张葆,杨扬.采用Hough变换的离焦模糊参数的估计[J].红外与激光工程,2012,41(10):2833-2837.
[3]姜明勇,陈向宁,喻夏琼.一种离焦模糊遥感图像盲复原方法[J].测绘科学,2012,37(4):135-137.
[4]梁敏,朱虹.基于边缘模糊频谱特征的散焦参数估计方法[J].计算机应用,2014,34(4):1177-1181.
[5]Jiang Y G,Wu Q,Guo P.Defocused image restoration using RBF network and Kalman filter[C].Proceedings of the 2005 IEEE International Conference on Systems,Man and Cybernetics,2005:2507-2511.
[6]Su L Y,Li F L,Xu F,et al.Defocused image restoration using RBF network and iterative Wiener filter in wavelet domain[C].Proceedings of the first International Congress on Image and Signal Processing.2008:311-315.
[7]Lee J,Fathi A S,Song S.Defocus Blur Estimation Using a Cellular Neural Network[C].Proceedings of the IEEE 12thInternational Workshop on Cellular Nanoscale Networks&Their Applications.2010:1-4.
[8]Moghaddam M E.A mathematical model to estimate out of focus blur[C].Proceedings of the 5th International Symposium on Image and Signal Processing and Analysis,2007:278-281.
[9]Moghaddam M E.A robust noise independent method to estimate out of focus blur[C].Proceedings of the IEEE International Conference on Acoustics,Speech,and Signal Processing.2008:1273-1276.
[10]杨亚威,耿志,王蕊,等.传统相机的空间变化离焦去模糊[J].电光与控制,2015,22(9):91-95.
[11]Gonzalez L,Velasco F,Angulo C,et al.A study on output normalization in multiclass SVMs[J].Pattern Recognition Letters,2013,34(3):344-348.
[12]李航.统计学习方法[M].北京:清华大学出版社,2012:96-135.
[13]Chang C C,Lin C J.LIBSVM:A library for support vector machines[EB/OL].http:∥www.csie.ntu.edu.tw/~cjlin/libsvm.