I have a set of RAW photos that I’d like to generate a linear color MSPECs for/correct exposure/white balance. Each photo has an 18% gray standard in the photo. For some photos, the MSPEC is successfully created and looks normal. For most others, however, the program flashes through images that look like typical outputs to end up with entirely white channels. I’ve tried editing min/max values but this hasn’t solved the problem. Thoughts?
Weird behaviour! Odd that some work and others don’t. With the ones that work does everything seem correct (e.g. when you measure the grey standard it’s showing 18% reflectance in all channels).
Are you using automatic dark correction? If so try turning that off and see what happens (i.e. with only the 18% standard).
With the white images, you could also go “image > adjust Brightness/contrast” and select auto. Do the image values make sense/can you see the image, and is the 18% standard 18%?
With the images that work, the reflectance is 18% in all channels — all good there. I tried the auto brightness, but it just turns the channels entirely blue/yellow/red, depending on which channel I was on when I set it.
Interestingly, although the MSPEC appears incorrect after initial creation, if I close the file and reload it, it is corrected (all channels reading 18). Weird, but we can work around it.
Another question for you — we are hoping to do a bunch of image analyses in R on our corrected photos. Can we somehow export our corrected photos as a typical image file for analysis?
How weird! If you work out how the bug is reproducible then do let me know. Otherwise, you can use “plugins > micaToolbox > Image Visualisation > Set Max and Min”, then set the range from 0-100 and this will apply the range to all channels.
For working with R, I would first suggest saving the image as a 32-bit linear TIFF and see if that works. i.e. load you mspec as a “Linear Colour” image and try that. Note that these images will be BIG (hundreds of MB each). If R only accepts 8-bit TIFFs (i.e. 8 bits per channel, also called 24 bit), then it gets difficult using linear images because they can’t preserve the full dynamic range. Avoid using JPG format, though it’s not the end of the world if you need this – as long as you’re consistent it should be fine.
Sadly R’s image libraries seem very sketchy with lots of broken dependencies on Linux, so I tend to try to do everything in ImageJ (which is much faster for image manipulation).