首页 | 本学科首页   官方微博 | 高级检索  
     

基于边缘预测和稀疏约束的湍流图像盲复原
引用本文:李晖晖 钱林弘 杨宁 胡秀华. 基于边缘预测和稀疏约束的湍流图像盲复原[J]. 仪器仪表学报, 2015, 36(4): 721-728
作者姓名:李晖晖 钱林弘 杨宁 胡秀华
作者单位:西北工业大学自动化学院;解放军93617部队
基金项目:航空基金(20131953022);中央高校基本科研业务费专项资金(3102014JCQ01062);装备研究基金(9140A06050113HK*****)项目资助
摘    要:大气湍流严重影响天文观测图像的成像效果,必须对退化图像进行处理才能获得清晰的图像。经典的湍流退化图像盲复原算法(IBD、NAS-RIF等)使用的先验知识过于简单,导致很多场合不能获得较优的复原效果。近几年提出的稀疏表达理论,使用自然图像边缘的稀疏先验信息指导图像复原,能复原出较多的细节,但它直接使用模糊图像的梯度图像指导点扩散函数复原,而模糊的梯度图像包含很多噪声和伪边缘,无效的梯度会误导点扩散函数的估计,从而使复原图像中出现较多伪迹。针对上述问题,提出了一种基于边缘预测和稀疏比值正则约束的湍流退化图像盲复原算法,该算法首先从当前的复原图像中预测出有效的边缘,然后将边缘预测信息与自然图像边缘的稀疏先验信息相结合指导点扩散函数复原,得到点扩散函数后,再通过一种非盲复原算法恢复出当前的目标图像,并将此复原图像作为下一次边缘预测的输入图像,如此迭代循环直到求出最终清晰的目标图像。所提算法结合了图像的先验信息与退化图像自身包含的有效信息,能有效抑制图像复原过程中产生的伪迹,获得令人满意的结果。针对多幅模拟的湍流退化图像进行仿真测试,验证了算法的有效性。

关 键 词:图像处理  湍流退化图像盲复原  稀疏正则约束  边缘预测  点扩散函数

Turbulence degraded image blind restoration based onsparsity regularization and edge prediction
Li Huihui;Qian Linhong;Yang Ning;Yang Weili;Hu Xiuhua. Turbulence degraded image blind restoration based onsparsity regularization and edge prediction[J]. Chinese Journal of Scientific Instrument, 2015, 36(4): 721-728
Authors:Li Huihui  Qian Linhong  Yang Ning  Yang Weili  Hu Xiuhua
Affiliation:Li Huihui;Qian Linhong;Yang Ning;Yang Weili;Hu Xiuhua;College of Automation,Northwestern Polytechnical University;The people’s liberation army 93617 troops;
Abstract:The image of astronomical observation is seriously affected by the atmospheric turbulence, which must be processed to obtain a clear one. The prior knowledge used in some classic blind restoration algorithm(IBD, NAS RIF, etc) is too simple to get a better recovery under some circumstances. Sparse representation proposed in recent years can restore more details guided by the prior sparse information of the edges in the natural images. However , the gradient of the blurred image which contains lots of noise and false edges is directly used to guide the recovery of the point spread function(PSF), and the invalid gradient will mislead the estimation of the PSF, and result in a lot of false edges in the restored image. To address the above problem, a method for blind image restoration based on edge prediction and sparse regular constraint is proposed. Firstly, effective edges are predicted from an estimated latent image. Then, the predicted edge information is combined with the prior information of the edges in the natural images to guide the restoration of PSF. Thirdly, through non blind deconvolution algorithm the latent image can be estimated. Fourthly, the recovered image is considered as the input of the next edge prediction. The above process is iterated until the final clear latent image is obtained. The proposed algorithm combines the effective information contained in the prior information of the natural images with that of the degraded images, which effectively suppress artifacts until more satisfactory results obtained. The effectiveness of the algorithm is verified through simulation of multiple turbulence degraded images
Keywords:
本文献已被 CNKI 等数据库收录!
点击此处可从《仪器仪表学报》浏览原始摘要信息
点击此处可从《仪器仪表学报》下载全文
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司  京ICP备09084417号