Created
August 1, 2012 17:05
-
-
Save robertskmiles/3228852 to your computer and use it in GitHub Desktop.
A script to automatically replace faces in group photos
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
#!/usr/bin/python | |
# face_replace.py | |
# Usage: python face_replace.py <image_file> [face index] | |
import sys | |
from opencv.cv import * | |
from opencv.highgui import * | |
from PIL import Image, ImageEnhance | |
import random | |
def detectObjects(image): | |
"""Converts an image to grayscale and returns the locations of any faces found""" | |
grayscale = cvCreateImage(cvSize(image.width, image.height), 8, 1) | |
cvCvtColor(image, grayscale, CV_BGR2GRAY) | |
storage = cvCreateMemStorage(0) | |
cvClearMemStorage(storage) | |
cvEqualizeHist(grayscale, grayscale) | |
cascade = cvLoadHaarClassifierCascade( | |
'/usr/share/opencv/haarcascades/haarcascade_frontalface_default.xml', | |
cvSize(1,1)) | |
scalefactor = 1.1 # How much to increase window size each pass | |
minsize = 50 # Smallest face to detect. Up this if you have small falsepositives | |
faces = cvHaarDetectObjects(grayscale, cascade, storage, scalefactor, 50, | |
CV_HAAR_DO_CANNY_PRUNING, cvSize(minsize, minsize)) | |
return [(f.x, f.y, f.x + f.width, f.y + f.height) for f in faces] | |
def meancol(source): | |
"""Find the mean colour of the given image""" | |
onepix = source.copy() | |
onepix.thumbnail((1,1),Image.ANTIALIAS) | |
return onepix.getpixel((0,0)) | |
def adjust(im, col, startcol=None): | |
"""Adjust the image such that its mean colour is `col`""" | |
if startcol is None: | |
startcol = meancol(im) | |
rband, gband, bband = im.split() | |
rbri, gbri, bbri = ImageEnhance.Brightness(rband), ImageEnhance.Brightness(gband), ImageEnhance.Brightness(bband) | |
rband = rbri.enhance((float(col[0]) / float(startcol[0]))) | |
gband = gbri.enhance((float(col[1]) / float(startcol[1]))) | |
bband = bbri.enhance((float(col[2]) / float(startcol[2]))) | |
im = Image.merge("RGB",(rband, gband, bband)) | |
return im | |
def main(): | |
inputfile = sys.argv[1] | |
image = cvLoadImage(inputfile) | |
print "Detecting faces..." | |
faceboxes = detectObjects(image) | |
print len(faceboxes), "faces found" | |
im = Image.open(inputfile) | |
mask = Image.open("circlemask.png") | |
if len(sys.argv) > 2: | |
faceindex = int(sys.argv[2]) | |
else: | |
print "Choosing random face" | |
faceindex = random.randrange(len(faceboxes)) | |
print "Using face", faceindex | |
newface = im.crop(faceboxes[faceindex]) | |
newfacemeancol = meancol(newface) | |
for box in faceboxes: | |
face = im.crop(box) | |
facemeancol = meancol(face) | |
adjustedface = adjust(newface, facemeancol, newfacemeancol) | |
size = (box[2] - box[0], box[3] - box[1]) | |
scalednewface = adjustedface.resize(size, Image.ANTIALIAS) | |
scaledmask = mask.resize(size, Image.ANTIALIAS) | |
im.paste(scalednewface, box, scaledmask) | |
im.show() | |
if __name__ == "__main__": | |
main() |
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment