傅里叶单像素成像技术研究进展(特邀)

计算机科学
作者
王天成 WANG Tiancheng,郁王涛 YU Wangtao,陈维芸 CHEN Weiyun,郭忠义 GUO Zhongyi
出处
期刊:Infrared and Laser Engineering [Shanghai Institute of Optics and Fine Mechanics]
卷期号:53 (9): 20240378-20240378 被引量:2
标识
DOI:10.3788/irla20240378
摘要

Significance In recent years, Fourier Single-Pixel Imaging (FSPI) technology has become a promising choice among the base-scanning Single-Pixel Imaging (SPI) methods due to its balance of high efficiency and high quality. This technology samples the target by projecting Fourier basis illumination patterns and achieves precise reconstruction from one-dimensional data to two-dimensional images with the help of efficient reconstruction algorithms. The FSPI technology does not rely on traditional array detectors and complex imaging systems, which doesn’t only reduce the hardware requirements for optical imaging but also broadens the range of applications, especially demonstrating significant advantages in resource-constrained or complex environmental scenarios. Therefore, research on the FSPI technology in the field of computational optical imaging holds important scientific value and application prospects. Progress Research on the FSPI technology contains some key aspects, such as light field modulation, path optimization, reconstruction algorithms, and application scenarios. In light field modulation, the spatial dithering strategy converts Fourier grayscale illumination fields into binary illumination fields, leveraging the high-speed modulation capabilities of spatial light modulators, but this approach sacrifices image spatial resolution, affecting reconstruction quality and clarity. Although the signal dithering strategy improves the modulation rate, it does not fully match the modulation rate of the spatial light modulator to some extents. As for the path optimization, traditional sampling paths can quickly recover targets and maintain good image quality, but their adaptability to different scenes is limited. Sparse sampling methods, based on signal sparsity, reduce the number of sampling points to lower data acquisition costs while retaining sufficient information for effective reconstruction; however, these methods often require prior knowledge and have high complexity. Adaptive sampling strategies offer targeted solutions by dynamically adjusting the position and density of sampling points to optimize imaging performance. Combining adaptive sampling with deep learning is an effective approach that can further optimize sampling strategies, however, it requires high pre-training costs and potentially suboptimal performance in different imaging systems. In reconstruction algorithms, Fourier inversion demonstrates robustness under specific noise conditions but may degrade image quality in practical environments, such as atmospheric turbulence. Second-order correlation imaging, which reconstructs scenes through statistical averaging, shows good anti-scattering performance but has slightly lower imaging efficiency. Compressed sensing algorithms use signal sparsity to reduce sampling rates, but their high computational complexity and time costs are still challenging. Deep learning methods significantly enhance imaging quality at low sampling rates but require large datasets for training, and discrepancies between training and real-world environments can affect overall accuracy. The FSPI technology has shown significant potential and great value in applications such as 3D imaging, edge detection, and scattering imaging, providing innovative solutions and new directions for future imaging technology developments. Conclusions and Prospects With its unique imaging principles and advantages, the FSPI technology plays an important role in many fields. However, there are still many unresolved problems and challenges. Future researches can focus on optimizing imaging algorithms, improving system stability, and expanding the range of applications. In particular, advances in adaptive light modulation and real-time processing will further enhance the capabilities of FSPI technology in complex and realistic imaging scenarios, thus promoting its transformation from theoretical research to practical applications.

科研通智能强力驱动
Strongly Powered by AbleSci AI
科研通是完全免费的文献互助平台,具备全网最快的应助速度,最高的求助完成率。 对每一个文献求助,科研通都将尽心尽力,给求助人一个满意的交代。
实时播报
zeee发布了新的文献求助10
刚刚
woshibyu发布了新的文献求助10
1秒前
1秒前
高大摇伽的应助被saturn采纳,获得10
1秒前
1秒前
Frieren完成签到 ,获得积分0
2秒前
Zhang完成签到,获得积分10
2秒前
隐形曼青的应助被Harvest采纳,获得20
2秒前
HH完成签到,获得积分10
2秒前
xiangfan完成签到,获得积分10
3秒前
科研通AI6.3的应助被wyyy采纳,获得10
3秒前
朴实依秋发布了新的文献求助10
3秒前
LanseR发布了新的文献求助10
3秒前
agegfsd发布了新的文献求助10
3秒前
4秒前
Sakura完成签到 ,获得积分10
4秒前
icebear完成签到 ,获得积分10
4秒前
斯文败类的应助被piccolovegeta采纳,获得10
5秒前
阿鑫完成签到 ,获得积分10
5秒前
123131发布了新的文献求助10
5秒前
东城区吴彦祖完成签到,获得积分10
5秒前
5秒前
Lemonhead的应助被ossc采纳,获得10
5秒前
充电宝的应助被可持续发展中采纳,获得30
6秒前
tczw667完成签到,获得积分10
6秒前
6秒前
ljh024发布了新的文献求助20
7秒前
renyun发布了新的文献求助10
7秒前
RW完成签到,获得积分10
7秒前
Jasper的应助被李金成采纳,获得10
7秒前
7秒前
陶醉恋风完成签到,获得积分10
7秒前
9秒前
可爱的函函的应助被CarryLJR采纳,获得10
9秒前
9秒前
10秒前
英俊的铭的应助被123131采纳,获得10
10秒前
LanseR完成签到,获得积分10
10秒前
Paeonolmite完成签到,获得积分10
10秒前
JamesPei的应助被满意莫茗采纳,获得10
11秒前
高分求助中
(应助此贴封号)【重要!!请各用户(尤其是新用户)详细阅读】【科研通的精品贴汇总】 10000
拉伸高分子熔体中缠结行为的模拟和小角中子散射研究 1700
Grapes 1000
The World Atlas of Wine 9th edition 1000
奈特基础组织学:配套组织病理学(第4版)(Netter's Essential Histology With Correlated Histopathology) 800
John McDowells Bildungstheorie 800
八旗制度の研究 600
热门求助领域 (近24小时)
化学 材料科学 医学 生物 纳米技术 计算机科学 工程类 有机化学 化学工程 生物化学 物理 复合材料 内科学 催化作用 光电子学 人工智能 细胞生物学 无机化学 物理化学 数学
热门帖子
关注 科研通微信公众号,转发送积分 8178375
求助须知:如何正确求助?哪些是违规求助? 9569182
关于积分的说明 21802781
捐赠科研通 7773247
什么是DOI,文献DOI怎么找? 3432799
关于科研通互助平台的介绍 2556793
邀请新用户注册赠送积分活动 2478346