CAJ | 학술논문

本文提出一种频域压缩多帧近视解卷积算法。首先，根据波前传感器获取的大气湍流相位估计短曝光图像的初始点扩展函数，并逐步将点扩展函数调整到准确形式。通过共轭梯度（CG ）算法交替解卷积频域代价函数估计目标图像和点扩展函数，为了有效压缩信息量，将图像频谱比增加到代价函数中，同时施加基本约束保证迭代快速收敛于全局最小解，采用结构自适应滤波器（SAAF ）达到消除噪声的同时保护重建图像细节。实验结果表明算法能够得到高质量的重建图像，性能优于所对比多帧Tikhonov正则化解卷积和多帧Richardson-Lucy的盲解卷积算法。
본문제출일충빈역압축다정근시해권적산법。수선，근거파전전감기획취적대기단류상위고계단폭광도상적초시점확전함수，병축보장점확전함수조정도준학형식。통과공액제도（CG ）산법교체해권적빈역대개함수고계목표도상화점확전함수，위료유효압축신식량，장도상빈보비증가도대개함수중，동시시가기본약속보증질대쾌속수렴우전국최소해，채용결구자괄응려파기（SAAF ）체도소제조성적동시보호중건도상세절。실험결과표명산법능구득도고질량적중건도상，성능우우소대비다정Tikhonov정칙화해권적화다정Richardson-Lucy적맹해권적산법。
A compact multiframe myopic deconvolution algorithm in the frequency domain is described that utilizes multi-frame short exposure images and the corresponding turbulent atmosphere phases from wave front sensing .Firstly ,the algorithm em-ploys the estimated point spread functions (PSFs ) based on the turbulent atmosphere phases to adjust to be better correct forms grad-ually .The cost function that uses Fourier spectral ratio to reducing the number of unknown variables is minimized alternately to re-store the object image and estimate PSFs by conjugate gradient algorithm .To diminish noise amplification in the frequency domain and avoid the reconstructed image turning to be a local minimum ,the structure-adaptive applicability filter (SAAF) and multiple ba-sis constraints are used .Computer simulations are conducted to investigate the performance of the proposed method offer the possi-bility of improved object reconstructions over multiframe Tikhonov regulation deconvolution and multiframe Richardson-Lucy decon-volution .