Is this histogram normal? Abnormal? Abbie Normal?

I shared over on CN an unstretched .tiff that was the output from pre-processing and stacking using Siril. One of the members there freaked out about the histogram being ‘quantized’. So I went back to a single original .fits file, opened it in Siril, and applied auto histogram stretch without doing anything else and got basically the same thing:
histo

This is from a QHY183C, captured using Ekos. I have checked the INDI driver settings and see nothing other than the expected 16-bit settings:
Image_info
Image_settings
I can talk myself into most any conclusion (in general, not just here). For example, the QHY183C is a 12-bit camera that upscales to 16-bit output by multiplying each pixel by 16. So in fact, there should only be a value for 1/16 of the 64k possibilities and this is pretty much expected, right? On the other hand, he seemed pretty freaked out by it.

So - if this is normal, I shall stop worrying. If it is abnormal, I presume the issue is on the Ekos side and will go ask in that forum.

For me, you have no reason to worry. Your image is 16 bit, it is normal to see this kind of histogram if you stretch it a lot.

All this historgram is showing is that you streched it too much. Typically this is leading to banding in the image. If you dont see banding, you are good to go. If you just worried about the historgam you can add a bit of (color) noise and et voilà :smiley:

Thanks for the confirmation. The CN user who raised the question went back to look at one of their own images from a 12-bit camera, and saw the same thing. So false alarm.

I also re-pre-processed a stack setting the preferences to 32-bit float instead of 16-bit unsigned, for the average stacking operations and that did substantially smooth out the auto-stretched histogram. Why didn’t I do that before? I guess my day job trained me to avoid creating “illusory” precision. Storing values that were measured as 4096 options in >4,000,000,000 bins sounded like over-precision.

Of course, the point is not to make pretty histograms. I am re-processing my current project with 32-bit float - we’ll see if I can notice a difference in the result. If nothing else, it might help stimulate the economy a bit by getting me to buy another SSD.

Abby Normal… :laughing:

Imgur