Hallo I have 2000 images and for each image I have extracted the 5 main colors (through a Python script), this data is stored in a xml file so that I can relate each image with its 5 main colors.
Now, how would one go about ordering the images by their main colors ? Apart from the vl technical perspective (OrderBy, etc.) I am not sure about the concept behind grouping or ordering palettes (spread of colors).
S.
as colors are inherently multidimensional this can be answered in many different ways.
- a simple way to start would be convert RGB to HSV and sort them by their hue.
- another option would be to really think of RGB values as a 3 dimensional position in a color cube and then find a path through the cube that always chooses the closest color to current color as the next color in the list
- (and many more)
1 Like
you could also take a look at other colour spaces like LAB for example. each of them maps differently and might be the right fit for your task.
1 Like
If this is presumed as an ontological problem I would suggest likewise to take in account quantities and qualities, like for instance these 5 color bands have all equal significance? And regarding quantity, do you store also the amount of pixels or a percentage of which colors are dominant? Additionally, these color bands have a constant distance in the Hue domain?
Otherwise, on @motzi track I would also try the Length (in case of non Lab colours - minimizing the dimensions to just one) and if you want more fun then maybe try K-Means / Centroids (and let the machine decide for you).