基于微分图像自相关的离焦模糊图像盲复原  

Blind Restoration for Defocus Blurred Image Based on Autocorrelation of Derivative Image

作　　者：赵琳[1] 金伟其[1] 陈翼男[1] 苏秉华[1] 

机构地区：[1]北京理工大学信息科学技术学院光电工程系,北京100081

出　　处：《光学学报》2008年第9期1703-1709,共7页Acta Optica Sinica

基　　金：中国信息安全产品测评认证中心开放基金(P2006019EF)资助课题

摘　　要：针对离焦模糊图像的盲复原算法的研究具有重要的实际意义和实用价值。根据光学离焦成像模型,研究提出了一种基于微分图像自相关的离焦模糊图像超分辨力盲复原算法,即首先采用拉普拉斯算子对离焦模糊图像进行二阶微分并求微分图像的自相关,然后从自相关结果所包含的信息中确定离焦模糊半径,最后以离焦模糊模型结合MPMAP超分辨力复原算法对离焦模糊图像进行盲复原。实验证明:算法能够以较高的精度估计出离焦模糊半径并实现离焦模糊图像的盲复原,该算法较其它同类算法在减少计算过程中需要考虑的各类因素的同时也减少了计算量,提高了结果精度,依靠超分辨力复原算法获取更多的复原图像信息,已在实际刑侦和物证鉴定的离焦模糊图像判读和鉴定中获得成功应用。The research on the blind restoration for the images blurred by defocus effect has important meaning and actual metric. A super-resolution blind restoration algorithm for defocus blurred images was suggested according to the model of optical defocusing by using autocorrelation of derivative image. The autocorrelation of the second-order derivative blurred image received by Laplacian was computed, the radius of defocusing blur was be confirmed via the information contained in the result of the autocorrelation, and the restoration was finally achieved by MPMAP super-resolution method using the estimated defocusing radius as parameter. Experiments illuminate that the proposed algorithm can accurately estimate the radius of defocusing blur and restore clear images. Compared with other methods, the proposed algorithm declines the computation quantity, improves result precision and obtaines more detailed information by using the super-resolution restoration algorithm. It has been successfully applied into judging or appraisal of defocus images in practice works.

关 键 词：图像复原 离焦模糊图像 自相关 拉普拉斯算子 超分辨力 

分 类 号：TN922.73[电子电信—通信与信息系统]