Measuring quality of scientific images is a highly relevant aspect of modern digital imaging. Especially in medical imaging, continuous quality testing is necessary to ensure diagnostic quality.
We show the types of measurements necessary to quantify and qualify the imaging performance of digital sensors embedded in modern cameras, medical imaging devices and other image acquisition devices. In principal measurements aiming at the spatial and frequency domain can be differentiated. In the spatial domain features like the characteristic curve, the dynamic range (DR), the homogeneity, the signal-to-noise ratio (SNR) and artifacts influence the reproduction. In the frequency domain the Modulation Transfer Function (MTF) and the Noise Power Spectrum (NPS), would describe the sharpness and the noise in relation to spatial frequencies.
In scientific imaging one usually wishes to draw conclusions from the displayed picture about the physical reality. For example, the correlation between pixel values and their corresponding photon flux in radiography. If the correlative function is known (usually the characteristic curve), one can transfer the image pixel values to corresponding physical parameters and ensure to measure the quality of the input signal. Desirably in general, this aspect is obligatory for testing medical image quality.
Some of the described artifacts can often be completely compensated (e.g. inhomogeneity) or sufficiently reduced (e.g. noise).
By means of plugins and programs developed at the Cologne University of Science we implemented the described measurements and describe ways to interpret them. This process includes methods for image improvement, like removing artifacts and correction of inhomogeneity.
image quality, image pre-processing, image enhancement, frequency domain, spatial domain
Presenting author: Holger Buhr
Organisation: Cologne University of Applied Sciences
co-authors: Christian Blendl