Benjamin_Loison/Robust_image_source_identification_on_modern_smartphones
1
0
Fork 0
Code Issues 68 Pull Requests Packages Projects Releases Wiki Activity

Generalize with a partial image SPLIT_N_X_N

Browse Source
This commit is contained in:
Benjamin Loison 2024-03-29 12:57:56 +01:00
parent 4bf9ef8206
commit 96bbd50a3b
Signed by: Benjamin_Loison
SSH Key Fingerprint: SHA256:BtnEgYTlHdOg1u+RmYcDE0mnfz1rhv5dSbQ2gyxW8B8
2 changed files with 25 additions and 10 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

19
algorithms/image_utils/image_utils.py
View File

@ -1,6 +1,7 @@
from PIL import Image
from PIL import Image, ImageFont, ImageDraw
import numpy as np
from matplotlib import pyplot as plt
import os
def randomGaussianImage(scale, size):
return np.random.normal(loc = 0, scale = scale, size = size)
@ -13,4 +14,18 @@ def showImageWithMatplotlib(npArray, title = None, cmap = 'viridis'):
plt.show()
def toPilImage(npArray):
return Image.fromarray(npArray)
return Image.fromarray(npArray)
def getPrnuShownAsSuch(size):
# Supports `WIDTH` > `HEIGHT` and conversely.
WIDTH, HEIGHT = size
TEXT = 'PRNU'
image = Image.new('L', size)
draw = ImageDraw.Draw(image)
# I guess that the maximal character height is higher than the maximal character width. Hence, the `TEXT` may not be spanned on the full width.
fontSize = min(HEIGHT, WIDTH // len(TEXT))
font = ImageFont.truetype(os.path.expanduser('~/.local/share/fonts/impact.ttf'), fontSize)
# Center vertically, especially in the case `HEIGHT` > `WIDTH`.
draw.text((0, HEIGHT // 2 - fontSize // 2), TEXT, 255, font = font)
return np.array(image)

16
datasets/noise_free_test_images/estimate_prnu.py
View File

@ -8,7 +8,7 @@ import sys
sys.path.insert(0, '../../algorithms/image_utils/')
from image_utils import showImageWithMatplotlib, randomGaussianImage, toPilImage
from image_utils import showImageWithMatplotlib, randomGaussianImage, toPilImage, getPrnuShownAsSuch
sys.path.insert(0, '../../algorithms/context_adaptive_interpolator/')
@ -25,15 +25,15 @@ datasetPath = 'no_noise_images'
# In addition to the range difference, note that the distribution in the first set of images was a Gaussian and here is very different and specific.
PRNU_FACTOR = 0.01
NOISE_FACTOR = 0.1
IMAGE_SIZE_SHAPE = (469, 704)
IMAGE_SIZE_SHAPE_4x4 = [size // 4 for size in IMAGE_SIZE_SHAPE]
SPLIT_N_X_N = 1
IMAGE_SIZE_SHAPE = [dimension // SPLIT_N_X_N for dimension in (704, 469)]
np.random.seed(0)
splitNxN = 4
#prnuNpArray = 255 * randomGaussianImage(scale = PRNU_FACTOR, size = IMAGE_SIZE_SHAPE)
prnuPil = Image.open(f'prnu_{"4x4_" if splitNxN == 4 else ""}noise.png').convert('F')
prnuNpArray = np.array(prnuPil) * PRNU_FACTOR
prnuNpArray = getPrnuShownAsSuch(IMAGE_SIZE_SHAPE) * PRNU_FACTOR
showImageWithMatplotlib(prnuNpArray)
def isIn256Range(x):
return 0 <= x and x <= 255
@ -46,8 +46,8 @@ for imageName in os.listdir(datasetPath):
imageWithoutPrnuPil = Image.open(imagePath).convert('F')
imageWithoutPrnuNpArray = np.array(imageWithoutPrnuPil)
m = imageWithoutPrnuNpArray.shape[0] // splitNxN
n = imageWithoutPrnuNpArray.shape[1] // splitNxN
m = imageWithoutPrnuNpArray.shape[0] // SPLIT_N_X_N
n = imageWithoutPrnuNpArray.shape[1] // SPLIT_N_X_N
imageWithoutPrnuNpArrayTiles = [imageWithoutPrnuNpArray[x : x + m, y : y + n] for x in range(0, imageWithoutPrnuNpArray.shape[0], m) for y in range(0, imageWithoutPrnuNpArray.shape[1], n)]
for imageWithoutPrnuNpArrayTile in imageWithoutPrnuNpArrayTiles: