def reduceWidth(im, energyImage):
mapV = cumulative_minimum_energy_map(energyImage, "VERTICAL")
seam = find_optimal_vertical_seam(mapV)
newIm = np.zeros((im.shape[0], im.shape[1] - 1, im.shape[2])).astype(int)
for i in xrange(len(seam)):
newIm[i] = np.delete(im[i], seam[i], 0)
return [newIm, energy_image(newIm)]
def reduceHeight(im, energyImage):
mapH = cumulative_minimum_energy_map(energyImage, "HORIZONTAL")
seam = find_optimal_horizontal_seam(mapH)
newIm = np.zeros((im.shape[0] - 1, im.shape[1], im.shape[2])).astype(int)
for i in xrange(len(seam)):
newIm[:, i] = np.delete(im[:, i], seam[i], 0)
return [newIm, energy_image(newIm)]
def reduceHeight(im, energyImage):
m, n = im.shape[: 2]
cumulativeMinimumEnergyMap = cumulative_minimum_energy_map(energyImage,'HORIZONTAL')
horizontal_seam = find_optimal_horizontal_seam(cumulativeMinimumEnergyMap)
reducedColorImage = np.zeros((m-1,n,3),dtype = np.uint8)
reducedEnergyImage = np.zeros((m-1,n),dtype = np.float64)
for col in range(n):
imageCopy = im[:,col,:]
mask = np.ones(m,dtype = bool)
mask[horizontal_seam[col]] = False
reducedColorImage[:,col,:] = imageCopy[mask]
reducedEnergyImage = energy_image(reducedColorImage)
return [reducedColorImage,reducedEnergyImage]
def reduceWidth(im, energyImage):
m, n = im.shape[:2]
cumulativeMinimumEnergyMap = cumulative_minimum_energy_map(
energyImage, 'VERTICAL')
verticalSeam = find_optimal_vertical_seam(cumulativeMinimumEnergyMap)
reducedColorImage = np.zeros((m, n - 1, 3), dtype=np.uint8)
reducedEnergyImage = np.zeros((m, n - 1), dtype=np.float64)
for rows in range(m):
imageCopy = im[rows, :, :]
mask = np.ones(n, dtype=bool)
mask[verticalSeam[rows]] = False
reducedColorImage[rows, :, :] = imageCopy[mask]
reducedEnergyImage = energy_image(reducedColorImage)
return [reducedColorImage, reducedEnergyImage]
def reduceHeight(im, energyImage):
input_color_image = im
input_image_nth_row, input_image_nth_col, input_image_nth_channel = input_color_image.shape
reducedColorImage = np.zeros((input_image_nth_row - 1, input_image_nth_col,
input_image_nth_channel),
dtype=np.uint8)
horizontalSeam = find_optimal_horizontal_seam(
cumulative_minimum_energy_map(energyImage, 'HORIZONTAL'))
for ith_col in xrange(0, input_image_nth_col):
reducedColorImage[:, ith_col,
0] = np.delete(input_color_image[:, ith_col, 0],
horizontalSeam[ith_col])
reducedColorImage[:, ith_col,
1] = np.delete(input_color_image[:, ith_col, 1],
horizontalSeam[ith_col])
reducedColorImage[:, ith_col,
2] = np.delete(input_color_image[:, ith_col, 2],
horizontalSeam[ith_col])
reducedEnergyImage = energy_image(reducedColorImage)
return reducedColorImage, reducedEnergyImage
from energy_image import energy_image
from reduceHeight import reduceHeight
import matplotlib.pylab as plt
import numpy as np
import scipy.misc
image = 'inputSeamCarvingPrague.jpg'
ei = energy_image(image)
image_array = np.asarray(plt.imread(image))
for i in range(0, 100):
image_array, ei = reduceHeight(image_array, ei)
scipy.misc.imsave('outputEnergyReduceHeightPrague.png', ei)
scipy.misc.imsave('outputReduceHeightPrague.png', image_array)
image = 'inputSeamCarvingMall.jpg'
ei = energy_image(image)
image_array = np.asarray(plt.imread(image))
for i in range(0, 100):
image_array, ei = reduceHeight(image_array, ei)
scipy.misc.imsave('outputReduceHeightMall.png', image_array)
# -*- coding: utf-8 -*-
"""
Created on Mon Sep 9 01:17:38 2019
@author: songl
"""
import matplotlib.pyplot as plt
from energy_image import energy_image
from reduceHeight import reduceHeight
pic1 = plt.imread('inputSeamCarvingPrague.jpg')
energy1 = energy_image(pic1)
for i in range(100):
pic1, energy1 = reduceHeight(pic1, energy1)
plt.imsave('outputReduceHeightPrague.png', pic1)
pic2 = plt.imread('inputSeamCarvingMall.jpg')
energy2 = energy_image(pic2)
for i in range(100):
pic2, energy2 = reduceHeight(pic2, energy2)
plt.imsave('outputReduceHeightMall.png', pic2)
bw = rgb2gray(im)
# use prewitt filters
kernely = np.array([[1, 1, 1], [0, 0, 0], [-1, -1, -1]])
# Perform x convolution
y = np.absolute(ndimage.convolve(bw, kernely))
# take threshold
threshold = np.mean(y)
y -= threshold
return y
### simple derivative
imgModify = mpimg.imread('./inputSeamCarvingPrague.jpg')
img = imgModify.copy()
modifiedEnergyImage = energy_image_simple(img)
energyImage = energy_image(img)
plt.imshow(energyImage, cmap=plt.get_cmap('gray'))
plt.show()
plt.imshow(modifiedEnergyImage, cmap=plt.get_cmap('gray'))
plt.show()
for i in xrange(100):
# progress(i, 100, status="reducing prague image width by 100 with simple one row derivative")
[imgModify, modifiedEnergyImage] = reduceWidth(imgModify,
modifiedEnergyImage)
[img, energyImage] = reduceWidth(img, energyImage)
plt.imshow(img)
plt.show()
plt.imshow(imgModify)
plt.show()
### take a high threshold cutoff
cum_eng_map[index_y][
index_x] = energyImage[index_y][index_x] + min(
cum_eng_map[index_y - 1][i],
cum_eng_map[index_y - 1][j],
cum_eng_map[index_y - 1][k])
if (seamDirection == "HORIZONTAL"):
cum_eng_map[:, 0] = energyImage[:, 0]
for index_x in range(1, width):
for index_y in range(0, height):
i = index_y - 1
j = index_y
k = index_y + 1
if (index_y == 0):
i = index_y
if (index_y + 1 == height):
k = index_y
cum_eng_map[index_y][
index_x] = energyImage[index_y][index_x] + min(
cum_eng_map[i][index_x - 1],
cum_eng_map[j][index_x - 1],
cum_eng_map[k][index_x - 1])
#plt.imshow(cum_eng_map)
#plt.show()
return cum_eng_map
ei = energy_image('inputSeamCarvingPrague.jpg')
#scipy.misc.imsave('EnergyPrague.png', ei)
map = cumulative_minimum_energy_map(ei, "VERTICAL")
#scipy.misc.imsave('EnergyMapPragueVertical.png', map)
import numpy as np
import math
from scipy.misc import imread, imsave
from scipy import ndimage
#from skimage import img_as_float
x1 = np.array([0, -(0.5), 0.5, 0])
x2 = np.array([0, 0.5, -0.5, 0])
xconvolution = ndimage.convolve(x1, x2, mode='constant', cval=0.0)
img = imread('inputSeamCarvingPrague.jpg')
#print img.dtype, img.shape
#plt.imshow(img)
#plt.imshow(img[: : ], cmap = plt.get_cmap('gray'))
#plt.imshow(energy_image(img), cmap = plt.get_cmap('gray'))
en_image = energy_image(img)
plt.figure()
plt.title('Gradient Energy Intensity')
plt.imshow(en_image)
plt.show()
v_cum_min_energy = cumulative_minimum_energy_map(en_image, 'VERTICAL')
plt.figure()
plt.title('Cumulative Vertical Intensity')
plt.imshow(v_cum_min_energy)
plt.show()
h_cum_min_energy = cumulative_minimum_energy_map(en_image, 'HORIZONTAL')
plt.figure()
plt.title('Cumulative Horizontal Intensity')
plt.imshow(h_cum_min_energy)
plt.show()
# -*- coding: utf-8 -*-
"""
Created on Sun Sep 13 09:41:15 2015
@author: sbhal
"""
from energy_image import energy_image
from reduceWidth import reduceWidth
import matplotlib.pyplot as plt
from scipy.misc import imsave
img = imread('inputSeamCarvingPrague.jpg')
print img.dtype, img.shape
en_im = energy_image(img)
for i in range(100):
img, en_im = reduceWidth(img, en_im)
plt.imsave("outputReduceWidthPrague", img)
img = imread('inputSeamCarvingMall.jpg')
print img.dtype, img.shape
en_im = energy_image(img)
for i in range(100):
img, en_im = reduceWidth(img, en_im)
plt.imsave("outputReduceWidthMall", img)
from energy_image import energy_image
from reduceHeight import reduceHeight
from progressBar import progress
import matplotlib.pyplot as plt
import matplotlib.image as mpimg
### prague image
imgPrague = mpimg.imread('./inputSeamCarvingPrague.jpg')
energyImagePrague = energy_image(imgPrague)
for i in xrange(100):
# progress(i, 100, status="reducing prague image height by 100")
[imgPrague, energyImagePrague] = reduceHeight(imgPrague, energyImagePrague)
plt.imshow(imgPrague)
# plt.savefig('./outputReduceHeightPrague.png')
plt.show()
### mall image
imgMall = mpimg.imread('./inputSeamCarvingMall.jpg')
energyImageMall = energy_image(imgMall)
for i in xrange(100):
# progress(i, 100, status="reducing mall image width by 100")
### reduce width
[imgMall, energyImageMall] = reduceHeight(imgMall, energyImageMall)
plt.imshow(imgMall)
# plt.savefig('./outputReduceHeightMall.png')
plt.show()
