先验与后验
计算机科学
概率逻辑
推论
正规化(语言学)
透视图(图形)
一致性(知识库)
波前
统计模型
相位恢复
航程(航空)
最大后验估计
算法
相(物质)
人工智能
反问题
机器学习
统计推断
计算复杂性理论
贝叶斯概率
理论计算机科学
冗余(工程)
不确定度量化
贝叶斯推理
最大似然
相位展开
物理系统
图形模型
相关性(法律)
数据挖掘
作者
Shuhe Zhang,Liangcai Cao
摘要
Recovering the phase of an optical wave is fundamental to understanding light–matter interaction and underpins a wide range of techniques in computational imaging, wavefront sensing, and optical metrology. Phase retrieval reconstructs the phase from intensity‐only measurements and has developed into diverse formulations. Despite their algorithmic differences, these approaches share a common goal: estimating the wavefield that best explains the observed measurements under physical and structural constraints. This review presents phase retrieval from the perspective of maximum a posteriori (MAP) estimation, using MAP as a statistical framework that connects measurement models, regularization principles, and algorithmic designs. Some methods explicitly minimize MAP objectives, whereas others are more appropriately understood as MAP‐inspired or conditionally related formulations that combine data consistency with prior information. Focusing on optical imaging, microscopy, and metrology, we systematically discuss the formulation of likelihood models, the role of priors, and their algorithmic realizations. The review provides a consolidated overview of developments, highlighting how probabilistic modeling bridges physical constraints and statistical inference. By clarifying this unified viewpoint, we outline current challenges and future directions toward more robust and interpretable computational imaging systems.
科研通智能强力驱动
Strongly Powered by AbleSci AI