Quantifying imaging imperfections resulting from nonideal fill factor and similar pixel-scale responsivity...

Image sensors can exhibit variations in the responsivity to light across the area of individual pixels. If the responsivity is uniform apart from a region with zero sensitivity, the situation is well described by the conventional concept of a fill factor, which gives the fraction of sensitive area in the pixel. In many practical cases, such responsivity variations are gradual, due to effects such as photon scattering or photoelectron diffusion between neighboring pixels. When characterizing the spatial responsivity of a complete camera, optical blur will tend to smear out pixel-scale variations. This paper generalizes the concept of a fill factor to cover such gradual sensitivity variations within a pixel, for an image sensor as well as for a complete camera. The core idea is that the effective fill factor (EFF) is the fill factor of a box-shaped pixel response that would produce the same maximum error in estimated reflectance as the actual camera response, when viewing a reflective scene. The resulting EFF is derived from the pixel sampling point spread function and sampling interval. It has a simple relation to an upper bound on the error in reflectance estimated from a reflected-light image. The EFF also has a simple relation to the average error in the total signal from a point source. It is noted that the spatial variation causing these image artifacts is specifically the variation of signal loss across the pixel. The EFF can supplement specification of many camera types, including thermal cameras as well as color and multispectral cameras based on filter arrays.