改进的基于白化滤波器的降晰函数尺寸估计方法  

Improved method of blur support size identification based on the whiten filter

作　　者：赵剡[1] 张春晓[1] 

机构地区：[1]北京航空航天大学仪器科学与光电工程学院,北京100083

出　　处：《红外与激光工程》2010年第5期896-901,共6页Infrared and Laser Engineering

基　　金：航空科学基金资助项目(20080112001)

摘　　要：给出了一种改进的降晰函数尺寸估计方法,它能为图像盲复原提供重要的先验信息,增强复原结果的可靠性。该算法首先由AR模型导出白化滤波器,然后对降质图像白化后的结果进行两种操作,以保证算法在低信噪比下的可行性:一是利用AR模型误差在边缘区域与平滑区域的差异,通过边缘提取与膨胀操作实现有效区域的快速提取;二是对有效区域通过平滑滤波进行噪声项的有效抑制。最后,对滤波后的有效区域进行相关值计算,依据取得最小值时的位移与降晰函数尺寸之间的关系估计降晰函数的尺寸。该方法经仿真实验验证:在低信噪比时,较改进前的方法能给出可靠的尺寸估计。An improved method of blur support size identification,which can provide the significant prior information for obtaining high-quality restored image in blind image restoration,was proposed.The method constructed a whiten filter derived from image AR model,and two measures were taken on the whiten image to ensure the feasibility of the method.One was to take advantage of the difference and obtain valid regions through edge detection and dilating operation,which was caused by the distribution of AR model error between the edge regions and the even regions.The other was to smooth the valid regions for suppressing the noise.Finally,the correlation values of smoothed valid regions were calculated,when the minimum was acquired,the blur support size could be estimated according to the relationship of the shift distance and the blur support size.The experimental results show that the method can estimate the more accurate blur support size in the low SNR condition compared with the original method.

关 键 词：降晰函数尺寸估计 图像盲复原 AR模型 白化滤波 相关运算 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]