摘要
将压缩传感理论引入超分辨率成像,得益于绝大多数图像在变换域中普遍具有稀疏性。在介绍压缩感知原理基础上,通过仿真分析表明,二维图像在变换域具有稀疏性;测量矩阵的性能越好重建图像效果越好;压缩感知采样仅用相当于传统图像30%测量值,就能恢复出与传统采样相当质量的图像;相对GPSR、GPSR+TV等算法,ADM算法超分辨率图像重建效果更佳。提出了一套基于4f系统的棱镜反射式压缩编码孔径光学成像系统,采用SLM作为编码模板完成对目标图像的调制和压缩,通过开发的基于全变分稀疏重建的ADM算法软件,实现了重建图像分辨率比CCD采集到的图像分辨率提高4倍的超分辨率重建效果。压缩感知成像技术解决了传统成像系统存在图像分辨率低、数据存储压力大、数据传输速度慢等问题,具有巨大应用潜力。
Compressive sensing theory is applied to super-resolution imaging for the general sparsity of most images.Compressive sensing principle and simulation show that two dimensional image in transform domain has sparsity.The better are the sensing matrix characteristics,the better is the image reconstruction effect.Only 30%measured value about traditional image’s is adopted by compressive sensing sampling,the image quality equal to that of traditional sampling is restored.Comparing with gradient projection sparsity reconstruction(GPSR)and GPSR+TV algorithms,the effect of super-resolution image reconstruction with alternating direction method of multipliers(ADM)algorithm is better.And an optical imaging system with prism reflective compressive code aperture is proposed based on 4f optical system,in which spatial light modulator(SLM)is used as coded template to modulate and compress target images.Four times super-resolution reconstruction effect than that of CCD is realized through developed ADM algorithm software based on total variation sparsity reconstruction.Traditional imaging problems such as low image resolution,high pressure data storage and slow data transmission are resolved by the compressive sensing imaging technology,which has great application potential.
作者
毕祥丽
许珈诺
BI Xiang-li;XU Jia-nuo(Science and Technology on Electro-Optical Information Security Control Laboratory,Tianjin 300308,China;School of Precision Instruments and Optoelectonics Engineering,Tianjin University,Tianjin 300072,China)
出处
《光电技术应用》
2018年第6期52-56,共5页
Electro-Optic Technology Application
关键词
压缩感知
图像稀疏性
压缩编码孔径成像
超分辨率重建
ADM算法
compressive sensing
image sparsity
compressive coded aperture imaging
super-resolution reconstruction
alternating direction method of multipliers(ADM)algorithm