Quote:

1. compute the average of the G and B tones
   
2. if the R tone is 1.5 times greater than that average, replace the R with the GB average value.
   

1. Make a selection on the eyes. It can/ should include the iris, but avoid skin tones. Feather a bit
   
2. Make a new layer from that: Ctrl-C, Ctr-V, Ctrl-Shift N. Let's call it "Eyes".
   
3. Make a copy of "Eyes", call it "Red 75%", start Color>Component>Channel mixer, set to "Monochrome", set Red to 75%, and Green and Blue to 0
   
4. Make a second copy of Eyes, call if G+B/2, start Color>Component>Channel mixer, set to "Monochrome", set Red to 0%, and Green and Blue to 50%
   
5. Make sure that all layers are invisible except these last two, with "G+B/2" over "Red 75%" and set G+B/2 to "Grain extract". The result should be eyes with a white disk where the red pupil was (with a dark spot where the white spot is).
   
6. Open the Channels list, and drag one of the R, G or B channels to the "main" list to copy it.
   
7. Back to the Layers list, make a last copy of "Eyes", call it "Correction", start Color>Component>Channel mixer, reset "Monochrome", set the Red channel to R=0%, G=50%, B=50%.
   
8. Make all layers invisible except the original picture and "Correction"
   
9. Right-click "Correction", and "Layer>Add layer mask". Initialize to "Channel" and pick the channel you copied at step 6. You should now have a fixed image.
   
10. At that point, you are editing the layer mask of "Correction" so you can tweak the amount of correction by playing with Levels or Curves (the most radical being thresholding the mask around 200, but this can make the fix limits a bit more visible).
    
    

(01-31-2018, 10:13 PM)Ofnuts Wrote: Editing in a photo is fraught with peril, in the good cases the subject looks weird, in the bad case s/he looks dead. But using the camera flash often makes it necessary to edit the picture to fix these awful red eyes, which is why I tried with much interest a technique alluded to on Pixls.us, for a new RawTherapee plugin:

Quote:

1. compute the average of the G and B tones
   
2. if the R tone is 1.5 times greater than that average, replace the R with the GB average value.
   

Fortunately, there is a Gimp equivalent:

1. Make a selection on the eyes. It can/ should include the iris, but avoid skin tones. Feather a bit
   
2. Make a new layer from that: Ctrl-C, Ctr-V, Ctrl-Shift N. Let's call it "Eyes".
   
3. Make a copy of "Eyes", call it "Red 75%", start Color>Component>Channel mixer, set to "Monochrome", set Red to 75%, and Green and Blue to 0
   
4. Make a second copy of Eyes, call if G+B/2, start Color>Component>Channel mixer, set to "Monochrome", set Red to 0%, and Green and Blue to 50%
   
5. Make sure that all layers are invisible except these last two, with "G+B/2" over "Red 75%" and set G+B/2 to "Grain extract". The result should be eyes with a white disk where the red pupil was (with a dark spot where the white spot is).
   
6. Open the Channels list, and drag one of the R, G or B channels to the "main" list to copy it.
   
7. Back to the Layers list, make a last copy of "Eyes", call it "Correction", start Color>Component>Channel mixer, reset "Monochrome", set the Red channel to R=0%, G=50%, B=50%.
   
8. Make all layers invisible except the original picture and "Correction"
   
9. Right-click "Correction", and "Layer>Add layer mask". Initialize to "Channel" and pick the channel you copied at step 6. You should now have a fixed image.
   
10. At that point, you are editing the layer mask of "Correction" so you can tweak the amount of correction by playing with Levels or Curves (the most radical being thresholding the mask around 200, but this can make the fix limits a bit more visible).
    
    

Before:


After:


The layers ("Visible" is extra, it is actually the result of step 5 (and what is copied to a Channel)):



Now, to adapt the technique to cats and dogs, whose eyes turn yellow or blue...

(02-05-2018, 03:16 PM)Issabella Wrote: Thank you so much Ofnuts for this new technique to repair this problem in some photos. I think It'll be very helpful. As soon as I find a photo with this problem, I'll use it.