摘要
采用频域多帧循环迭代解卷积算法(CIBD),针对提高复原图像的准确性和快速性两个方面进行研究。以退化序列中任意帧作为起始帧,逐次增加迭代帧,确保更多的观测帧参与循环迭代解卷积以增加复原的准确性;通过图像间的相关矩阵估计初始点扩展函数(PSF),采用尺度梯度投影法,自适应迭代步长,增加迭代终止条件等措施提高算法的收敛速度。实验结果表明,采用提议的算法能够有效地重建不同大气湍流条件下的远距离观测图像,性能优于传统多帧盲反卷积(MBD)迭代算法。
A circulate iterative blind deconvolution （CIBD） algorithm to restore turbulence-degraded images in the frequency domain is described, which focus on enhancement of accuracy and rapidity. The accuracy is enhanced by beginning with random frame and gradually increasing new frame to circulate iteration to get satisfactory restoration. Some measures are implemented to improve convergence rate, such as estimated point spread functions （PSF） by correlation matrix, adaptive steplength updating rule, condition of iterative termination and scaled projection gradient strategy. The experimental results of simulation show that the algorithm is efficient to reconstruct the degraded images from remote ground-based observations under different atmospheric turbulence intensities, and its performance is better than traditional multiframe blind deconvolutions （MBDs）.
出处
《激光与光电子学进展》
CSCD
北大核心
2014年第7期41-48,共8页
Laser & Optoelectronics Progress
基金
安徽省高校省级自然科学项目(KJ2013B052
KJ2013B065)
关键词
大气光学
盲反卷积
循环迭代
多尺度投影梯度
自适应迭代步长
atmospheric optics
blind deconvolution
circulate iteration
scaled projection gradient
adaptive steplength
