去模糊
黑森矩阵
共焦
对数
计算机科学
图像复原
光学
人工智能
反褶积
盲反褶积
点扩散函数
图像处理
物理
计算机视觉
图像(数学)
数学
数学分析
应用数学
作者
Tao He,Yasheng Sun,Jin Qi,Haiqing Huang,Jie Hu
出处
期刊:Applied Optics
[Optica Publishing Group]
日期:2019-06-20
卷期号:58 (19): 5148-5148
摘要
Confocal fluorescence microscopy has become a cardinal workhorse instrument in biological research due to its high imaging speed and tissue penetration depth. Unfortunately, the sampled fluorescence signals are intrinsically distorted by optical blurs and photon-counting noise, and the deconvolution method has been introduced to attenuate these degradations. In this paper, we focus mainly on scenarios suffering from severe noise due to low exposure time in a fast-imaging system. To begin with, a Hessian penalty was adopted to depress the artificial staircase effects that were caused by the first-order model (e.g., total variation). Then, to compensate for the weak ability to remove blurring and the produced white-point artifacts of the second-order penalty, we additionally proposed a consistent constraint along the temporal axis based on structural continuity. A remarkable merit of the spatiotemporal fused regularization is retaining the ability of the Hessian matrix to keep details smooth while effectively removing blurring. We employed an alternating-direction-method-of-multipliers algorithm for the corresponding optimization problem. Finally, we conducted experimental comparisons of both the simulated and practical confocal platform, and the excellent performance of the proposed approach reflects the efficiency of the confocal deconvolution work.
科研通智能强力驱动
Strongly Powered by AbleSci AI