import cv2

from paint_mask import MaskPainter

from move_mask import MaskMover

from poisson_image_editing import poisson_edit

from skimage.transform import match_histograms

from skimage.filters import unsharp_mask

#import argparse

import getopt

import sys

from os import path

import numpy as np

from color_transfer import color_transfer

def usage():

\t-m\t(Optional) Specify a mask image with the object in white and other part in black, ignore this option if you plan to draw it later.")

def split_chan(im, sz):

return dark, bright, r, g, b

def get_mask(img):

return imgf

def resize(img):

small = cv2.resize(img, (0, 0), fx=0.25, fy=0.25)

return small

if __name__ == '__main__':

# parse command line arguments

args = {}

try:

opts, _ = getopt.getopt(sys.argv[1:], "hs:t:m:p:")

except getopt.GetoptError as err:

# print help information and exit:

print(err) # will print something like "option -a not recognized"

print("See help: main.py -h")

exit(2)

for o, a in opts:

if o in ("-h"):

usage()

exit()

elif o in ("-s"):

args["source"] = a

elif o in ("-t"):

args["target"] = a

elif o in ("-m"):

args["mask"] = a

else:

assert False, "unhandled option"

#

if ("source" not in args) or ("target" not in args):

usage()

exit()

#

source = cv2.imread(args["source"])

target = cv2.imread(args["target"])

b,g,r = cv2.split(target)

cv2.imwrite('red.png', r)

cv2.imwrite('blue.png', b)

cv2.imwrite('green.png', g)

# source = cv2.bilateralFilter(source, 11, 81, 81)

# cv2.imwrite('bilateral.png', source)

# source = color_transfer(source=target, target=source)

# source = cv2.resize(source, (0, 0), fx=0.5, fy=0.5)

# target = cv2.resize(target, (0, 0), fx=0.5, fy=0.5)

# for i in range(3):

# source[:,:,i] = cv2.equalizeHist(source[:,:,i])

# cv2.imwrite('equ.png', source)

# clahe = cv2.createCLAHE(clipLimit=2.0, tileGridSize=(8, 8))

# source = clahe.apply(source)

# cv2.imwrite('clahe.png', source)

source = cv2.fastNlMeansDenoising(source, None, 3, 7, 25)

cv2.imwrite('denoise_nlm.png', source)

# source = cv2.medianBlur(source, 11)

# cv2.imwrite('denoise_median.png', source)

#

# source = unsharp_mask(source, radius=3, amount=3, multichannel=True)

# source = np.uint8(source * 255)

# cv2.imwrite('usm.png', source)

# kernel = np.array([[-1, -1, -1], [-1, 9, -1], [-1, -1, -1]])

# source = cv2.filter2D(source, -1, kernel)

# cv2.imwrite('sharpen.png', source)

# source = match_histograms(source, target, multichannel=True)

# cv2.imwrite('color_match.png', source)

# cv2.imshow("denoised", source)

# cv2.imshow("target", target)

# cv2.waitKey()

if source is None or target is None:

print('Source or target image not exist.')

exit()

# if source.shape[0] > target.shape[0] or source.shape[1] > target.shape[1]:

# print('Source image cannot be larger than target image.')

# exit()

# draw the mask

mask_path = ""

if "mask" not in args:

print('Please highlight the object to disapparate.\n')

mp = MaskPainter(args["source"])

mask_path = mp.paint_mask()

else:

mask_path = args["mask"]

# adjust mask position for target image

# print('Please move the object to desired location to apparate.\n')

# mm = MaskMover(args["target"], mask_path)

# offset_x, offset_y, target_mask_path = mm.move_mask()

target_mask_path = args["mask"]

# blend

print('Blending ...')

# target_mask = cv2.imread(target_mask_path, cv2.IMREAD_GRAYSCALE)

target_eye_mask = cv2.imread(target_mask_path, cv2.IMREAD_GRAYSCALE)

offset = (128, -34)

#

target_mask = np.uint8(get_mask(target) * 255)

cv2.imwrite('target_mask_dark_chan_original.png', target_mask)

kernel = np.ones((21, 21), np.uint8)

target_mask = cv2.dilate(target_mask, kernel, iterations=1)

cv2.imwrite('target_mask_dark_chan.png', target_mask)

target_mask = target_mask * target_eye_mask

cv2.imwrite('target_mask_final.png', np.uint8(target_mask * 255))

# offset = offset_x, offset_y

poisson_blend_result = poisson_edit(source, target, target_mask, offset)

cv2.imwrite(path.join(path.dirname(args["source"]), 'target_result.png'),

poisson_blend_result)

print('Done.\n')