MTF measurements, identifying bias, and estimating uncertainty

The modulation transfer function (MTF) describes how an imaging system will modify the spatial frequency content of a scene. Many performance metrics are strongly dependent on the MTF as it provides information of the limiting resolution. To measure the MTF, an image of a known or assumed scene is analyzed. In this correspondence we will detail potential issues that can contribute uncertainty or bias into the calculation of the MTF when using a tilted edge to super-resolve the system MTF of a sampled imaging system. Differences in measuring the system MTF can be categorized into 4 categories: data corruption, equipment, operator selection, and system under test effects. For each category we provide notional examples to demonstrate the severity of the measurement uncertainty as well as best practices to avoid or reduce their influence. We provide the full 2D derivation of the tilted edge technique, highlighting the impact of non-uniformity. We will discuss the influence of finite regions of interest (ROI) and stray-light, defective pixels, non-uniform illumination, and non-square pixels. Additionally we show how confidence intervals from sensor noise can be estimated and how they are related to frame averaging and ROI size. In support of the reproducible research effort, the Matlab functions associated with this work can be found on the Mathworks file exchange [1].