Digitization bit depth in Fluoview 3.1

[glenmac]

LOCI seems to be not reading the digitization bit depth in Olympus OIF files created with Fluoview 3.1
z-stacks are being opened with the histograms set to the min/max of first slice. I can supply an OIF file or a complete dataset.

Thanks,
Glen

[rleigh]

We can try to reproduce and fix the issue if you could supply an OIF file. Ideally, we'd need all the other accompanying files like the .files directory as well--you could just zip them all up together. You can upload them here:

http://qa.openmicroscopy.org.uk/qa/upload/


Many thanks,
Roger Leigh

[glenmac]

Thanks Roger, I just uploaded MacDonaldFV3.1OIF.zip. 3 channel z-series, 12-bit data range, from one of my users. Channel 1 has no saturated pixels in the first few slices, Channel 2 has none at all, channel 3 has some saturated speckles in the first 1 slices, with some later slices containing saturated cell bodies. So Ch. 3 is the only channel that scales correctly.

The LOCI plugin was able to correctly read the dig depth in the OIF and OIB formats at one point. I haven't paid attention to the issue so I don't know when it stopped working or whether the issue is Bio-formats or Fluoview.
Regards
Glen

[rleigh]

Got the file now, thanks.

[rleigh]

Hi Glen,

From what I can see, while the image planes are 12-bit data, they are accompanied by 12-bit lookup tables of 16 bit values:

Indexed = true (false color, 16-bit LUT: 3 x 65536)

So after doing the LUT lookups, the pixel data is all 16-bit. This is the reason for the altered bit depth, at least. Regarding the autoscaling, I'll need to check exactly where in the code this occurs: bio-formats, the imagej LOCI plugin, or imageJ itself.


Regards,
Roger

[glenmac]

HI,
Here is a snippet of an exchange I had with Curtis, November, 2010. I just checked the OIF file for BitsPerPixel and ValidBitCounts. Both are still "12".
Me:>I just poked around the metadata on the Olympus OIF and OIB. Both the Metadata option and the OME_XML >options display only that the filetype is "uint16". Here is a snippet I found in the OIF file that seems to >indicate bit-depth in the 'ValidBitCounts=12'. There is probably a similar statement in the OIB format.

Curtis: I checked, and the Bio-Formats FV1000 reader does indeed parse the ValidBitCounts metadata appropriately. However, the parsed bit depth is never explicitly shown in either metadata window. I added this to the latest trunk build, so now when you check "Display metadata" it will show a BitsPerPixel entry at the top that indicates the number of valid bits. Even if the PixelType is uint16, it should still show 12 for the BitsPerPixel for your data—please let us know if not. It should also scale accordingly when the Autoscale option is unchecked.

Glen

[rleigh]

Hi Glen,

To clarify, is the problem here primarily the autoscaling not working?

I loaded your sample image into FiJi and imported using the LOCI importer. This is shown as 16-bit pixel data, but with a 12 bit range in the B&C controls. I imported both with and without the "autoscale" option enabled. The min/max values are as follows:

Code: Select all

Channel  Image data  No autoscale  Autoscale
1        291/4095    0/4095        71/4095
2        210/2295    0/4095        73/2295
3        142/4095    0/4095        73/4095


The image data is the min/max of a max intensity z projection, so it's the min/max of all pixels for each channel.

Given the saturation in channels 1 and 3, the max value isn't changed for these. I'm not entirely sure why the min values aren't better scaled. I also found an issue with autoscaling not working when you import a restricted z range.

Does this match what you observed? If this is not what you were expecting, what was the desired behaviour here?

Thanks,
Roger

[glenmac]

I went back and collected my own images to control the intensities.
With Autoscale ON and opening the image as merged, the histogram is set to min/max of the entire stack. - that is the expected behavior. When importing the stack split into channels, the histogram for each channel is set to the min/max of the initial slice. - that seems to be inappropriate behavior, since it leads to saturated display in the brighter image planes. Autoscale should use the histogram from entire stack for channel being opened.


Glen

[rleigh]

I can reproduce this with split channels (and also split focal plane). I opened a ticked for this (#10092). You should receive a CC for this.

Regards,
Roger