Problem with Unmixcolors and pixel intensity scales


A few weeks ago it was suggested to me that the Unmixcolors module would be better for histology, since it is proper to process each staining type that there is as option. At first I liked it a lot, because for that particular region it solved the problem well. However, I noticed that there is a strong modification of the pixel intensity scale of the image, which changes everything in terms of high throughput processing.
Attached there’s a series of images exemplifying this modification in the pixel intensity, comparing the processed output to a grayscaled and inverted image, along with its pixel intensity histograms. Apparently, there is a sort of saturation of the DAB processed image in a way that too many pixels get to max intensity and renders particles less identifiable. In that sense, just gray-scaling them and and inverting them seem to result in an image closer to the original both visually and in the histogram (see attached).

Is there a way to correct or avoid this saturation (I really don’t know if I’m using this term correctly)?

Thanks in advance,


Hi Cesar,

Sorry we missed this post. Have you found a satisfactory explanation? I’m afraid we would need more information to give you a good answer. Namely what would help would be the pipeline. I’m just not sure what manipulations are happening in what order (Morph-Invert?, UnmixColors, etc)


Hi David,

To those images, I just loaded the original images as color RGB, and used unmixcolors module and evaluated their pixel intensity OR I just loaded the images as color RGB and used colortogrey module and looked the pixel intensity scale.

I did not put the other parts of the pipeline bc this is the focus of my problem. I have been using the second process so far, or taking the pictures in B/W. But seeing what is happening with the unmixcolor module and its saturation seemed an important issue to be reported.