I posted it in the original description which was deleted for some reason. First I created the Granger rainbow images for each dichromacy using a tool named Color Oracle. Then I used K-means clustering to get a reduced 8 color palette for each one of them. I applied the reduced palettes back to the corresponding Granger rainbow images to get the areas of these colors. I then layered the results to get the common areas. The colors of the palette were picked from around the centroids of these areas without any further normalization.Note that some of the final areas turned to get very small so most normalization impairs people with a certain dichromacy. I failed with two attempts to normalize this palette with some HSV/HSL+Grayscale pattern and decided to publish the raw results instead.