OpenCV vs Labview Images Greyscale (U16) - Difference in values

[alexderjuengere · Accepted solution]

ok, this is happening already when the images are loaded 

notice, the whole "stitching" is happening trough the "tunnel mode: concatenate" - this has no effect at all.

I have no idea why this is happening, but there must be some kind of image-processing in the imaq sub.vis involved.

I guess its some process step either in imaq read or imaq imagetoarray - both subvis are wrapping dll calls ... so, I can't dig deeper

[kirkland77]

So, IMAQ is incorrect for rescaling these values, as when we read images, we shouldn't have the pixel values changes automatically. I have tried to determine if there is some arbitrary threshold that IMAQ uses to decided to scale, however, I was not able to narrow anything down. Based of the histogram it appears to use a factor of 130,000 to scale the values down but that could be subjective to this image. So, should I use Python instead of IMAQ when trying to join images in LabView, due to its unpredictable behavior?

[alexderjuengere]

I can picture ( 😉 ) two use-cases:

the imaq images "looks" better, but the  python-opencv imaqs is more "acurate"

[kirkland77]

My thoughts as well. Cheers @alexderjuengere 😀

[alexderjuengere]

I had to look at this again ... there is obviously some histogram matching - with reference to the first image - going on in imaq.

https://en.wikipedia.org/wiki/Histogram_matching

I tried a quick attempt to reproduce the histogram matching, but I probably just use imaq, if I ever need this.

this might be an actual use case for those tensorflow neural nets ... nevermind.

[alexderjuengere]

maybe I need this again