Image texture is the term given to the information-bearing fluctuations such as those for skin, grass and fabrics. Since
image processing aimed at reducing unwanted fluctuations (noise are other artifacts) can also remove important texture,
good product design requires a balance between the two. The texture-loss MTF method, currently under international
standards development, is aimed at the evaluation of digital and mobile-telephone cameras for capture of image texture.
The method uses image fields of pseudo-random objects, such as overlapping disks, often referred to as ‘dead-leaves’
targets. The analysis of these target images is based on noise-power spectrum (NPS) measurements, which are subject to
estimation error. We describe a simple method for compensation of non-stationary image statistics, aimed at improving
practical NPS estimates. A benign two-dimensional linear function (plane) is fit to the data and subtracted. This method
was implemented and results were compared with those without compensation. The adapted analysis method resulted in
reduced NPS and MTF measurement variation (20%) and low-frequency bias error. This is a particular advantage at low
spatial frequencies, where texture-MTF scaling is performed. We conclude that simple trend removal should be used.