sliding_histogram_wally.py

import matplotlib.pyplot as plt
from skimage.io import imread
from skimage.morphology import disk
from skimage.filters import rank
import numpy as np

im_all = imread("/home/scott/Dropbox/wally/ww2.jpg")
all_red = im_all[:, :, 0]
im_wally = imread("/home/scott/Dropbox/wally/wally2.png")
wally_red = im_wally[:, :, 0]

def windowed_histogram_similarity(image, selem, reference_hist, n_bins):
    # Compute normalized windowed histogram feature vector for each pixel
    px_histograms = rank.windowed_histogram(image, selem, n_bins=n_bins)
    X = px_histograms
    Y = reference_hist
    num = (X - Y) ** 2
    denom = X + Y
    denom[denom == 0] = np.infty
    frac = num / denom
    chi_sqr = 0.5 * np.sum(frac, axis=2)
    similarity = 1 / (chi_sqr + 1.0e-10)
    return similarity

# need to reduce grayscales values to 16 levels
def quant_16(image):
	image_out = image // 16
	return image_out

wally16 = quant_16(wally_red)
all16 = quant_16(all_red)

# compute wally histogram and normalize
wally_hist, _ = np.histogram(im_wally.flatten(), bins = 16, range = (0, 16))
wally_hist = wally_hist.astype(float) / np.sum(wally_hist)

selem = disk(25)

similarity = windowed_histogram_similarity(all16, selem, wally_hist,
                                           wally_hist.shape[0])

fig = plt.figure(figsize = (12,12))
ax1 = plt.subplot(2,1,1)
ax1.imshow(im_all)
ax1.set_title("Original Image")
ax2 = plt.subplot(2,1,2)
ax2.imshow(all16, cmap = plt.cm.gray)
ax2.imshow(similarity, cmap = plt.cm.Spectral_r, alpha = 0.5)
ax2.set_title("Similarity to Wally")
plt.savefig("/home/scott/Dropbox/wally/histogram_window.png")