Project Summary

Takes an image and outputs a "color scheme" of the most prominent colors in an image

Motivation

I have often looked at images and wished I could get a color palette of the most prominent colors so I could channel that color scheme in a painting or other design project so I set out to create a program that does just that

Code Example (Clustering Algorithm)

# weighting colors that make up over 10% of the image less

# initializing vars
cutPix = []
dict = {}

# loopinng through all the pixels
for p in allPix:
    
    # creating a dictionary with all the pixels and the number of times
    # they appear in allPix
    if p in dict:
        dict[p]+=1
    else:
        dict[p] =1

# the cutOff for a "large" amount of a color is if it makes up about over
# 20% of all the pixels
cutOff = int(len(allPix)*(20/100))

for k in dict:
    
    # if the count of the pixel is greater than the cutoff
    if(dict[k] > cutOff):
        # add that pixel 2% less than it appears in allPix, or in other words
        # only include 98% of the iterations of that pixel
        for i in range(int(0.98*dict[k])): cutPix.append(k)
    else:
        # otherwise add it the original amount of times it appears in allPix
        for i in range(dict[k]): cutPix.append(k)


# loop through all the pixels
for p in cutPix:
    # we will start with the max distance and max pix values
    minDist = pixDist(255,255,255,0,0,0)
        
    for c in range(len(centroids)):
        r1,g1,b1 = centroids[c]
        r2,g2,b2 = p
        
        # find the pixels distance to the current centroid
        dist = pixDist(r1,g1,b1,r2,g2,b2)
        
        # if the pix distance from the centroid < the min distance btwn
        # that pix and the other centroids then make that dist the minDist
        # and that centroid the closest centroid
        if(dist < minDist):
            minDist = dist
            index = c # represents index of cluster list OR index of which centroid closest to
            
    # add that pixel to the closest centroid
    clusters[index].append(p)
        
    # change the centroid to average together 
    r1,g1,b1 = centroids[index]
    r2,g2,b2 = p
    
    # taking avg of old centroid and new cluster addition to
    # get new centroid
    centroids[index] = (r1+r2)//2,(g1+g2)//2,(b1+b2)//2   

Output Examples

Credits

Colors dataset used for color names is a subset of a dataset taken from https://data.world/dilumr/color-names

Yeshiva University © Juliet Nasar