Vignetting

Darker corners can come from three places. The optics themselves deliver less light off axis: the further a point is from the center, the more obliquely its light passes through the system, and some is cut off by the edges of lenses and stops. Illumination that is off center or set up wrongly adds its own falloff, which Köhler illuminationLoading... removes. And a camera whose sensor is larger than the image circle the camera adapter delivers loses its corners outright, often as a sharp dark ring rather than a gradual fade.

Only the first is unavoidable. The second is fixed with two adjustments, the third with the right adapter. What remains is handled by flat-field correctionLoading....

Measure it on blank glass as the ratio of corner brightness to center brightness. In brightfield, the shortfall reads directly as optical densityLoading...: corners at 85% of the center add log₁₀(1 ÷ 0.85) = 0.07 OD in every channel, and corners at 70% add 0.155 OD.

That offset is applied to everything in the corner, so identical cells measure as more strongly stained there than in the center. Weakly stained cells near a positivity threshold can cross it on position alone. In a whole-slide scan the error is not random: the same cells would score differently if the scan grid had fallen a few hundred micrometers to one side.

A whole-slide scan is built from hundreds or thousands of fieldsLoading..., each carrying the same falloff toward its edges. StitchedLoading... together, the pattern repeats as a regular grid: bright tile centers, darker seams. Blending at the seams can soften it to the eye, but it does not remove the underlying error; it only spreads it.

The grid is also a convenient diagnostic. If a stitched blank or lightly stained region shows tile-shaped structure, the flat fieldLoading... is missing or stale. One reference per objective and setup corrects every tile at once, because the falloff is a property of the optics, not of the specimen.