Just wanted to check on here before going to Github, but I’m noticing the colour picker for Color EQ is not acting how I would expect. I would expect the intersection of the crosshairs would pinpoint the exact colour to sample. But there seems to be an offset.
The crosshairs/reticle is clearly over the yellow area but the colour being picked up is the nearby orange. This is on a PNG file with no other modules activated (other than default), but I first noticed it on raw files. It does not seem to be related to cropping or perspective.
I believe it is using the whole circle for sampling when I just tested it. Because your circle crosses over two colors the result you get is expected in my view. However, I could see an advantage in being able to reduce or increase the size of the circle and I wonder if that is possible?
it used to be the case that dt picks colours from the full-image-low-res preview pipeline. the pleasures of low-res+cropped dual pipelines. since the picker is required for a thing or two, you don’t want it to disappear once you pan it out of view.
I see the issue at lower zooms too, and regardless, sometimes you want to zoom in to pick an exact colour.
Yep, that first one was mine back in the day. If I recall it was verified as undesirable and due to be fixed, but not sure if it ever was.
So, I’m left thinking this may be “as designed” but not particularly desirable. I don’t see the point in having crosshairs if they aren’t accurate. I think I’ll open a Github issue.
When you zoom in that much depending on your display more scaling is going on so I think you are really only color “accurate” for point estimates at the pixel level at 100 percent. At least I feel like that and then if the point picker does maybe a 3x3 or something to select the color scaling might also impact it…I guess you would have to dive in to the code and see how the selection is defined??
Thanks for that. When you open it please let me know so I can add my support to the issue. It would be very useful to be able to make the size of the circle bigger or smaller like we do with the circular drawn masks.
@europlatus I have added to your github post. In a reply to your github post @Jens-Hanno_Schwalm said only a single pixel was read and not an area, but in the sample given in the OP here the color appears to be the average of an area but maybe the single pixel is taken from the area where there is a blending of the two colors. I certainly feel there is room to improve this tool’s accuracy and I would propose DT adopting a system like GIMP’s color picker where you can adjust the radius of the ample average. It can be set to 1 if the user wants but being able to sample a larger area may be more appropriate. The method used in DT’s color calibration or exposure module might provide a suitable approach as the selected area is easily adjusted in these two modules.
While playing around with this color picker issue on a very recent build of DT … WIndows 11…
I noted that you can still see things like this issue with the GF state in the Color Eq that was closed… and you see it in this case with or with out the HQ preview and with or without OpenCL…
Enabling the Color Eq in this image creates some desaturation and adds creates what appears like CA… disabling the module or the GF removes it…no sliders are moved…
