Auto White Balance

Auto white balance has no reference to look at, so it assumes one. The commonest rule is gray world: the average of the scene is neutral, so the camera scales its channels until the mean red, green and blue are equal. A second rule, white patch, assumes the brightest area is white. Many cameras combine the two, and most apply the result continuously, recalculating on every frame or every few frames.

For an everyday photograph, a varied scene of many colors, the gray-world average is often close to neutral and the guess works. A microscope field is not a varied scene: it is dominated by one or two stains, so the assumption is wrong before the first frame.

Take a field that averages R 150, G 120, B 110, a typical brown-and-blue tissue field. Gray world scales the channels to their common mean of about 127: gains of ×0.84, ×1.06 and ×1.15. Blank glass that was already balanced at 230, 225 and 228 becomes 194, 238 and 263, so blue now clipsLoading... and the glass turns blue-green.

Every color in the field shifts the same way, partly removing the stain the camera was supposed to record. And because the correction depends on how much tissue is in view, it differs from field to field: a sparse field and a dense field of the same slide get different gains, and the same cells measure differently depending on their neighbors.

Switch the camera to manual or fixed white balance and set it on blank glass (see white balanceLoading...). Then check it held: image a field of blank glass and a field crowded with tissue, and compare the glass in both. With a fixed balance, the glass reads the same in each; with auto still active, it shifts between them.

Some cameras and acquisition programs restore their default mode when they restart or when a profile is loaded, so make the check part of starting a session. Save the white-balance mode with the acquisition settings, not just the gains.