Some key issues raised by MELK, with audience participation

The MELK approach to immuno-fluorescence, brilliantly developed by Walter Schubert, produces many aligned images of a biological specimen, each image staining for a different biomolecule. We (that is, the audience and the speaker) will discuss various problems that arise when processing MELK images and approaches to their solution.
a) Noise removal. Many MELK images show a considerable amount of noise which isn't really noise, but rather unwanted signal. This is fluorescence caused by "non-specific binding". What techniques are available for removing, or at least reducing, this unwanted signal? This is related to an attempt to make immuno-fluorescence into a quantitative matter (see b) below). Current practice among biologists is to binarize the signal---either the protein is there or it's not---but that approach discards a lot of the information.
b) Normalization of images. Suppose you take an immuno-fluorescent picture of two different sections related to the same stage of the same underlying biological process. In general, the "scales" on which one sees the two images will differ, probably in a non-linear way. What methods are there of changing scale in each image, so that the two images become more directly comparable?
c) We look at Insulin on the SAME section, first at the 4th MELK cycle, then at the 35th MELK cycle. How do these relate---quantitatively?
d) Segmentation of MELK images into cells. This problem is much more difficult with cell sections where the cells are contiguous, than it is with in vitro material where the cells are usually separate.
e) What kinds of image viewing software is needed for immuno-fluorescence in general, and for MELK in particular. This may be related to b) above.