Brettschneider JA, Warnett JW, Nichols TE and Kendall WS
Higher level spatial analysis of dead pixels on detectors based on local grid geometry
Abstract: Let a collection Bi (i ∈ I) of binary values indexed by a grid I ⊆ Z2 be a model to describe dysfunctional pixels in images obtained by a detector based technology. While there are obvious global measures to quantify the amount of damage, natural questions also arise around the spatial distribution of the dysfunctional pixels and how observed patterns of dysfunctional pixels may be interpreted.
After modelling occurrences of dysfunctional pixels as a planar point process we develop a higher level approach for analysing their spatial distributions. Key idea is to move from the notion of a dysfunctional pixel to the concept of a damage event defined by configurations of dysfunctional pixels using a typology based on local grid geometry. High density regions can be detected using density estimation of the damage event process, so remaining areas becomes suitable candidates for complete spatial randomness. This approach decouples observed damage from the detector resolution prescribed by I and from the exact shape of dysfunctional pixel configurations.
We proposes a detector quality toolkit that allows users to monitor their technology following these principles. The methods allow users of detector based imaging technologies to detect, distinguish and monitor different types of quality damage and to identify the ones linked to specific causes. We apply our methods to a collection of bad pixel maps obtained as part of regular monitoring routines of a detector used in X-ray computed tomography.