红外与毫米波学报, 2017, 36(4): 498, 网络出版: 2017-04-01
AExperimental study for coding computational imaging and quality evaluation of reconstructed image
1中国科学院上海技术物理研究所 中科院空间主动光电技术重点实验室, 上海 200083
2中国科学院大学, 北京 100049
计算成像 压缩感知 图像评价 信噪比估计 子空间分析 computational imaging compressed sensing image quality evaluation signal-to-noise (SNR) estimation subspace analysis
针对国产化大面阵红外器件技术缺乏, 用红外焦平面探测器获取图像时难以消除自身的非均匀性以及信噪比低, 航空航天成像应用中的图像采集、传输、存储成本越来越高等问题.论文引入了像面编码计算成像, 首先分析了这种成像系统的原理模型, 然后将压缩感知理论应用于计算成像中, 搭建成像原理样机, 进行非压缩和压缩的计算成像实验.最后在重构的图像质量评价中引入了信号子空间分析方法, 进行重构图像的信噪比估计, 实验结果表明, 这种信噪比估计方法更加准确有效, 并且可以以此作为实验结论给出实际压缩成像时需要的合理采样次数.
Infrared focal plane array detector suffers from problems including the lack of localization technology of infrared devices, intrinsic non-uniformity, and low signal-to-noise ratio (SNR). The costs for image acquisition, transmission, and storage became higher and higher for the applications of aerospace imaging. To overcome these difficulties, this paper analyzed the principle of computational imaging system, and the compressive sensing theory was introduced for imaging. Then the imaging prototype was built, and compressed and uncompressed imaging had been investigated. Finally, the method of the signal subspace analysis was introduced for quality evaluation of the reconstructed image, and the SNR of the reconstructed image was estimated using this method. Experimental results demonstrate that the SNR estimation method is accurate and effective, which can give a reasonable number of samples required for the actual compressed imaging.