研究目的
To develop a high-precision estimate algorithm for zero-frequency noise power spectrum (NPS) in digital imaging systems, addressing the susceptibility to distortion and low precision in conventional methods.
研究成果
The proposed compensated partial-sum algorithm significantly improves the precision of zero-frequency NPS estimates, achieving up to 14 times better precision than conventional methods. It provides unbiased estimates with faster convergence, making it suitable for practical applications in digital radiography imaging, though parameter D must be optimized based on extensive experiments for various detectors.
研究不足
The algorithm's accuracy depends on the choice of segment size D, which requires empirical determination for different detectors. The method assumes rapid decay of spatial correlation, which may not hold for all imaging systems. Overlap in subimages can slightly degrade performance compared to non-overlap cases.
1：Experimental Design and Method Selection:
The study involved designing a compensated partial-sum algorithm for estimating zero-frequency NPS, based on theoretical models from signal processing and radiography imaging standards. The method includes using synthetic data for initial validation and experimental data from a radiography detector.
2：Sample Selection and Data Sources:
Synthetic data were generated for numerical analysis, and practical x-ray images were acquired from a CsI(Tl)-scintillator-based flat panel radiography detector under RQA 5 condition, following IEC62220-1 standards.
3：List of Experimental Equipment and Materials:
A flat panel radiography detector (model not specified, brand DRTECH Co. Ltd.), with a pixel pitch of 140 μm/pixel and gray level resolution of 14 bits/pixel, was used. X-ray images were the primary data source.
4：Experimental Procedures and Operational Workflow:
For synthetic data, Monte Carlo simulations with 1,000 experiments were conducted. For experimental data, subimages of size D x D were constructed from acquired images, and the proposed algorithm was applied to compute zero-frequency NPS estimates with compensation constants. Overlap and non-overlap cases were considered to vary sample size M.
5：Data Analysis Methods:
Relative precision was calculated as the standard deviation of estimates to its mean ratio. Estimates were compared between the proposed method, conventional IEC standard, and extrapolation methods using statistical analysis and graphical comparisons.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
