Components of noise at high and medium frequencies in digital x‐ray images: Precise determination of the quantum noise fraction

Noise in digital x-ray imaging systems comes from three sources: quantum, electronic, and structural, which impact image quality and diagnostic accuracy. These noise components are typically derived from the terms of a second-degree polynomial that models the variance of image noise as a function of the entrance air kerma at the detector. However, this method suffers from significant uncertainties due to the presence of low-frequency trends in the images, which require restricting the analysis to a small region. This study focuses on the precise determination of the quantum noise fraction (QNF) in radiological imaging. For this purpose, the uniform images are processed before performing the noise component decomposition to extract the medium and high frequencies of the image and eliminate low-frequency trends. Using an orthogonal wavelet packet transform, noise decomposition is performed in the wavelet domain with a focus on the diagonal components in medium and high-frequency sub-bands. Different flat-panel detectors from various manufacturers were studied, producing uniform images acquired under controlled conditions using an RQA5 beam quality. Results demonstrated that the wavelet domain provides much lower uncertainty in QNF estimation compared to the image domain. This reduction in uncertainty enables more reliable comparisons between different detectors and imaging conditions. Furthermore, QNF was found to be higher in mid-frequency bands than in high-frequency bands at lower exposure levels. These findings highlight the importance of precise noise characterization in optimizing detector performance and improving radiological image quality. The proposed wavelet-based methodology offers a robust approach to noise decomposition, allowing for more accurate assessments of imaging system capabilities and guiding improvements in digital radiography.