Histogram matching in python

Inspired by ths question on stackoverflow.

The author wants to make the colors in one picture similar to the colors in another. Here is his source picture:

He wants to get the tint from that picture:

and apply it to the source.

The algorithm I suggest gives the following result:

before: after:

The algorithm is called histogram matching and essentially means applying histogram equalization to both pictures, and then creating the pixel value translation function from the two equalization functions.

Let's use the code from Jan Erik Solem as a base.

   def histeq(im,nbr_bins=256):

   #get image histogram
   imhist,bins = histogram(im.flatten(),nbr_bins,normed=True)
   cdf = imhist.cumsum() #cumulative distribution function
   cdf = 255 * cdf / cdf[-1] #normalize

   #use linear interpolation of cdf to find new pixel values
   im2 = interp(im.flatten(),bins[:-1],cdf)

   return im2.reshape(im.shape), cdf

And modify it so it becomes a histogram matching function working with both single-channel and multichannel(like RGB) images:

from scipy.misc import imsave, imread
import numpy as np

imsrc = imread("source.jpg")
imtint = imread("tint_target.jpg")

nbr_bins=255
if len(imsrc.shape) < 3:
    imsrc = imsrc[:,:,np.newaxis]
    imtint = imtint[:,:,np.newaxis]

imres = imsrc.copy()
for d in range(imsrc.shape[2]):
    imhist,bins = np.histogram(imsrc[:,:,d].flatten(),nbr_bins,normed=True)
    tinthist,bins = np.histogram(imtint[:,:,d].flatten(),nbr_bins,normed=True)

    cdfsrc = imhist.cumsum() #cumulative distribution function
    cdfsrc = (255 * cdfsrc / cdfsrc[-1]).astype(np.uint8) #normalize

    cdftint = tinthist.cumsum() #cumulative distribution function
    cdftint = (255 * cdftint / cdftint[-1]).astype(np.uint8) #normalize


    im2 = np.interp(imsrc[:,:,d].flatten(),bins[:-1],cdfsrc)



    im3 = np.interp(im2,cdftint, bins[:-1])

    imres[:,:,d] = im3.reshape((imsrc.shape[0],imsrc.shape[1] ))

try:
    imsave("histnormresult.jpg", imres)
except:
    imsave("histnormresult.jpg", imres.reshape((imsrc.shape[0],imsrc.shape[1] )))

Written on May 18, 2015