Department of Dental Hygiene, College of Health Science, Gachon University, 191, Hambakmoero, Yeonsu-gu, Incheon, Republic of Korea.
Department of Radiation Convergence Engineering, Yonsei University, 1, Yonseidae-gil, Wonju-si, Gangwon-do, Republic of Korea.
Microsc Microanal. 2020 Oct;26(5):929-937. doi: 10.1017/S143192762000183X.
This study aimed to develop and evaluate a blind-deconvolution framework using the alternating direction method of multipliers (ADMMs) incorporated with weighted L1-norm regularization for light microscopy (LM) images. A presimulation study was performed using the Siemens star phantom prior to conducting the actual experiments. Subsequently, the proposed algorithm and a total generalized variation-based (TGV-based) method were applied to cross-sectional images of a mouse molar captured at 40× and 400× on-microscope magnifications and the results compared, and the resulting images were compared. Both simulation and experimental results confirmed that the proposed deblurring algorithm effectively restored the LM images, as evidenced by the quantitative evaluation metrics. In conclusion, this study demonstrated that the proposed deblurring algorithm can efficiently improve the quality of LM images.
本研究旨在开发和评估一种基于交替方向乘子法（ADMMs）的盲反卷积框架，并结合加权 L1 范数正则化用于明场显微镜（LM）图像。在进行实际实验之前，使用西门子星型幻影进行了预仿真研究。随后，将所提出的算法和基于全广义变分（TGV-based）的方法应用于在 40×和 400×显微镜放大倍数下捕获的小鼠磨牙的横截面图像，并对结果进行比较，比较了得到的图像。模拟和实验结果均证实，所提出的去模糊算法可以有效地恢复 LM 图像，定量评估指标也证明了这一点。总之，本研究表明，所提出的去模糊算法可以有效地提高 LM 图像的质量。
