基于噪声分析和稀疏正则化的图像盲复原方法
|
摘要
在海事搜救过程中,机载红外相机拍摄的红外图像由于直升机振动、气流扰动、高速飞行以及红外相机摆扫等因素,严重影响图像质量.根据直升机载红外相机成像特点,提出了一种基于噪声分析和稀疏正则化的图像盲复原方法.该方法首先分析了成像过程中的噪声分布,并对噪声进行预处理,再根据稀疏表达理论,用图像边缘的稀疏先验信息指导点扩散函数复原,接着通过非盲复原方法得到目标图像,将目标图像作为下一次迭代的输入图像,如此循环迭代得到清晰图像.最后,对仿真模糊图像和实拍模糊图像进行了复原实验.实验结果表明这种方法能有效改善图像质量,并且在处理实拍运动模糊图像时,相比其他复原方法效果更好.
In the course of maritime search and rescue,infrared image captured by helicopter airborne infrared camera has a poor image quality because of the helicopter vibration,air turbulence,high speed flight and infrared camera sweeping. According to the imaging characteristics of the helicopter airborne infrared camera,a blind restoration method for blurry images based on noise analysis and sparsity regularization was proposed. Firstly,noise distribution in the imaging process is analyzed and the noise is pre-processed. Then,according to the sparse representation theory,sparse prior information of the edges in the images is used to guide the restoration of PSF. After that,we can obtain the target image through non-blind method. The target image will be used in the next iteration. The iteration will not end until a clear image is obtained. Experiments were performed both on simulated blurry images and real blurry images. Experimental results show that our method can effectively improve the image quality.Compared with other methods,our method has a better effect on real blurry images.
In the course of maritime search and rescue,infrared image captured by helicopter airborne infrared camera has a poor image quality because of the helicopter vibration,air turbulence,high speed flight and infrared camera sweeping. According to the imaging characteristics of the helicopter airborne infrared camera,a blind restoration method for blurry images based on noise analysis and sparsity regularization was proposed. Firstly,noise distribution in the imaging process is analyzed and the noise is pre-processed. Then,according to the sparse representation theory,sparse prior information of the edges in the images is used to guide the restoration of PSF. After that,we can obtain the target image through non-blind method. The target image will be used in the next iteration. The iteration will not end until a clear image is obtained. Experiments were performed both on simulated blurry images and real blurry images. Experimental results show that our method can effectively improve the image quality.Compared with other methods,our method has a better effect on real blurry images.
引文
[1]BAI Xiang-En,YI Jia.Parameter estimation method for infrared motion-blur images[J].Transducer and Microsystem Technologies(白响恩,易佳.红外运动模糊图像参数估计方法.传感器与微系统),2015,34(9):27-29.
[2]Yu-Li You,M.Kaveh.A Regularization Approach to Joint Blur Identification and Image Restoration[J].IEEE Transactions on Image Processing,1996,5(3):416-428.
[3]Tony F.Chan,Chiu-Kwong Wong.Total Variation Blind Deconvolution[J].IEEE Transactions on Image Processing,1998,7(3):370-375.
[4]Rob Fergus,Barun Singh,Aaron Hertzmann et al.Removing Camera Shake from a Single Photograph[J].ACM Transactions on Graphics,2010,25(3):787-794.
[5]Dilip Krishnan,Terence Tay,Rob Fergus.Blind Deconvolution Using a Normalized Sparsity Measure[J].IEEE Computer Vision and Pattern Recognition,2011,42(7):233-240.
[6]XIAO Su,ZHENG Ying.Sparse representation based multiregularized image deconvolution[J].Computer Engineering and Applications(肖宿,郑颖.基于稀疏表示的多正则优化图像复原.计算机工程与应用),2015,51(12):203-207.
[7]LIU Cheng-Yun,CHANG Fa-Liang.Blind moving image restoration based on sparse representation and Weber's law[J].Optics and Precision Engineering(刘成云,常发亮.基于稀疏表示和Weber定律的运动图像盲复原.光学精密工程),2015.2,23(2):600-608.
[8]LI Hui-Hui,QIAN Lin-Hong,YANG Ning,et al.Turbulence degraded image bilind restoration based on sparsity regularization and edge prediction[J].Chinese Journal of Scientific Instrument(李晖晖,钱林弘,杨宁,等.基于边缘检测和稀疏约束的湍流图像盲复原.仪器仪表学报),2015.4,36(4):721-728.
[9]GUO Cong-Zhou,QIN Zhi-Yuan.Blind restoration of nature optical images based on non-convex high order total variation regularization[J].Optics and Precision Engineering(郭从洲,秦志远.非凸高阶全变差正则化自然光学图像盲复原.光学精密工程),2015.12,23(12):3490-3499.
[10]LI Jian-Bing,MA Ren,LI Dan-Yang,et al.Tikhonov and total variation regularization method for blind image restoration[J].Journal of Nanjing University of Posts and Telecommunications(Natural Science Edition)(李俭兵,马忍,李丹阳等.基于Tikhonov和全变分正则化混合约束盲去模糊方法.南京邮电大学学报(自然科学版)),2016.6,36(3):68-73.
[11]WANG Bin-Hui,XU Yong-Yong.Test for the normality on the probability plots and its application[J].Mathematical Statistics and Applied Probability(王斌会,徐勇勇.正态性检验的图示方法及其应用.数理统计与应用概率),1996.9,11(3):249-255.
[12]LI Yu-Mei.The methods for testing normal distrbution of data[J].Journal of Huaihua Uniersity(李玉梅.数据的正态性检验方法.怀化学院学报),2015.11,34(11):81-82.
[13]Amir Beck,Marc Teboulle.A Fast Iterative ShrinkageThresholding Algorithm for Linear Inverse Problems[J].SIAM on Imaging Sciences,2009,2(1):183-202.
[14]Anat Levin,Rob Fergus,Frédo Durand et al.Image and Depth from a Conventional Camera with a Coded Aperture.ACM Transactions on Graphics,2007,26(3):70.
[2]Yu-Li You,M.Kaveh.A Regularization Approach to Joint Blur Identification and Image Restoration[J].IEEE Transactions on Image Processing,1996,5(3):416-428.
[3]Tony F.Chan,Chiu-Kwong Wong.Total Variation Blind Deconvolution[J].IEEE Transactions on Image Processing,1998,7(3):370-375.
[4]Rob Fergus,Barun Singh,Aaron Hertzmann et al.Removing Camera Shake from a Single Photograph[J].ACM Transactions on Graphics,2010,25(3):787-794.
[5]Dilip Krishnan,Terence Tay,Rob Fergus.Blind Deconvolution Using a Normalized Sparsity Measure[J].IEEE Computer Vision and Pattern Recognition,2011,42(7):233-240.
[6]XIAO Su,ZHENG Ying.Sparse representation based multiregularized image deconvolution[J].Computer Engineering and Applications(肖宿,郑颖.基于稀疏表示的多正则优化图像复原.计算机工程与应用),2015,51(12):203-207.
[7]LIU Cheng-Yun,CHANG Fa-Liang.Blind moving image restoration based on sparse representation and Weber's law[J].Optics and Precision Engineering(刘成云,常发亮.基于稀疏表示和Weber定律的运动图像盲复原.光学精密工程),2015.2,23(2):600-608.
[8]LI Hui-Hui,QIAN Lin-Hong,YANG Ning,et al.Turbulence degraded image bilind restoration based on sparsity regularization and edge prediction[J].Chinese Journal of Scientific Instrument(李晖晖,钱林弘,杨宁,等.基于边缘检测和稀疏约束的湍流图像盲复原.仪器仪表学报),2015.4,36(4):721-728.
[9]GUO Cong-Zhou,QIN Zhi-Yuan.Blind restoration of nature optical images based on non-convex high order total variation regularization[J].Optics and Precision Engineering(郭从洲,秦志远.非凸高阶全变差正则化自然光学图像盲复原.光学精密工程),2015.12,23(12):3490-3499.
[10]LI Jian-Bing,MA Ren,LI Dan-Yang,et al.Tikhonov and total variation regularization method for blind image restoration[J].Journal of Nanjing University of Posts and Telecommunications(Natural Science Edition)(李俭兵,马忍,李丹阳等.基于Tikhonov和全变分正则化混合约束盲去模糊方法.南京邮电大学学报(自然科学版)),2016.6,36(3):68-73.
[11]WANG Bin-Hui,XU Yong-Yong.Test for the normality on the probability plots and its application[J].Mathematical Statistics and Applied Probability(王斌会,徐勇勇.正态性检验的图示方法及其应用.数理统计与应用概率),1996.9,11(3):249-255.
[12]LI Yu-Mei.The methods for testing normal distrbution of data[J].Journal of Huaihua Uniersity(李玉梅.数据的正态性检验方法.怀化学院学报),2015.11,34(11):81-82.
[13]Amir Beck,Marc Teboulle.A Fast Iterative ShrinkageThresholding Algorithm for Linear Inverse Problems[J].SIAM on Imaging Sciences,2009,2(1):183-202.
[14]Anat Levin,Rob Fergus,Frédo Durand et al.Image and Depth from a Conventional Camera with a Coded Aperture.ACM Transactions on Graphics,2007,26(3):70.