Department of Oral Surgery, University Medical Center of the Johannes Gutenberg University Mainz, Germany.
Dentomaxillofac Radiol. 2011 May;40(4):222-9. doi: 10.1259/dmfr/33029984.
The aim of this study was to illustrate the influence of digital filters on the signal-to-noise ratio (SNR) and modulation transfer function (MTF) of digital images. The article will address image pre-processing that may be beneficial for the production of clinically useful digital radiographs with lower radiation dose.
Three filters, an arithmetic mean filter, a median filter and a Gaussian filter (standard deviation (SD) = 0.4), with kernel sizes of 3 × 3 pixels and 5 × 5 pixels were tested. Synthetic images with exactly increasing amounts of Gaussian noise were created to gather linear regression of SNR before and after application of digital filters. Artificial stripe patterns with defined amounts of line pairs per millimetre were used to calculate MTF before and after the application of the digital filters.
The Gaussian filter with a 5 × 5 kernel size caused the highest noise suppression (SNR increased from 2.22, measured in the synthetic image, to 11.31 in the filtered image). The smallest noise reduction was found with the 3 × 3 median filter. The application of the median filters resulted in no changes in MTF at the different resolutions but did result in the deletion of smaller structures. The 5 × 5 Gaussian filter and the 5 × 5 arithmetic mean filter showed the strongest changes of MTF.
The application of digital filters can improve the SNR of a digital sensor; however, MTF can be adversely affected. As such, imaging systems should not be judged solely on their quoted spatial resolutions because pre-processing may influence image quality.
本研究旨在说明数字滤波器对数字图像的信噪比（SNR）和调制传递函数（MTF）的影响。本文将讨论可能有助于降低辐射剂量生成具有临床实用价值的数字射线照片的图像预处理。
测试了三种滤波器，即算术平均滤波器、中值滤波器和高斯滤波器（标准差（SD）= 0.4），核大小分别为 3×3 像素和 5×5 像素。创建了具有确切高斯噪声量的合成图像，以收集应用数字滤波器前后 SNR 的线性回归。使用具有定义线对每毫米数量的人工条纹图案来计算应用数字滤波器前后的 MTF。
具有 5×5 核大小的高斯滤波器可实现最高的噪声抑制（SNR 从合成图像中测量的 2.22 增加到滤波图像中的 11.31）。最小的噪声减少是用 3×3 中值滤波器发现的。中值滤波器的应用不会导致不同分辨率下的 MTF 发生变化，但会导致较小结构的删除。5×5 高斯滤波器和 5×5 算术平均滤波器显示出最强的 MTF 变化。
数字滤波器的应用可以提高数字传感器的 SNR；然而，MTF 可能会受到不利影响。因此，成像系统不应仅根据其标称空间分辨率进行判断，因为预处理可能会影响图像质量。
