最大似然空间变化图像恢复算法  

Space variant image restoration based on maximum likelihood

作　　者：王治乐[1] 赵明[1,2] 李博[3] 孟祥龙[3] 智喜洋[1] 

机构地区：[1]哈尔滨工业大学空间光学工程研究中心,黑龙江哈尔滨150001 [2]大连海事大学信息科学与技术学院,辽宁大连116026 [3]中国白城兵器试验中心,吉林白城137001

出　　处：《红外与激光工程》2012年第7期1947-1951,共5页Infrared and Laser Engineering

基　　金：国家自然科学基金(61007008)

摘　　要：在光学图像处理中,把光学成像系统看做线性空间变化系统具有普遍意义。从实际光学系统成像过程出发,考虑光学系统的点扩散函数的空间变化特性和探测器噪声特性,建立了空间变化成像模型。在此成像模型基础上,基于最大似然法提出了空间变化的Richardson-Lucy(SVRL)图像恢复算法。为了分析SVRL算法的性能,实验中利用ZEMAX软件计算不同视场的点扩散函数,而后利用此空间变化点扩散函数进行成像仿真,得到仿真成像结果,最后分别采用0视场、0.5视场、0.7视场、1视场的点扩散函数以及空间变化点扩散函数对仿真图像进行恢复。实验结果表明,对于实际的空间变化光学系统,SVRL算法的图像恢复效果十分有效。For optical image processing, the imaging system is space-variant, and thus space-variant image restoration was a state-of-art and challenging problem. Practical imaging process of optical system was analyzed, and then space-variant imaging model was constructed by considering the space-variant point spread function of optical system and the noise characteristic of detector. With the space-variant imaging model, a space-variant image restoration algorithm was proposed based on the maximum likelihood method, it was called space-variant Richardson-Lucy （SVRL） image restoration algorithm. To analyze the performance of SVRL, space variant PSF （SVPSF） was generated by ZEMAX, simulated imaging was performed based on imaging equation with SVPSF. By comparing the restoration results by PSFs of 0, 0.5, 0.7, 1.0 FOV and SVPSF, it was proved that SVRL algorithm is more effective than space-invariant restoration result.

关 键 词：图像恢复 空间变化图像恢复 SVRL算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]