Wiener filter #10

New Issue

Closed

opened 2024-03-20 12:00:33 +01:00 by Benjamin_Loison · 3 comments

Benjamin_Loison commented 2024-03-20 12:00:33 +01:00

Owner

Copy Link

Copy Source

https://web.archive.org/web/20240320105334/https://web.stanford.edu/class/ee368/Handouts/Lectures/2014_Spring/8-Linear-Image-Processing/Wiener_Filtering.pdf#page=5

pdfimages Wiener_Filtering.pdf -png -f 5 -l 5 Wiener_Filtering

Unclear how to get Wiener_Filtering-006.png from Wiener_Filtering-001.png.

from scipy.signal import wiener
import matplotlib.pyplot as plt
import numpy as np
from PIL import Image

img = Image.open('Wiener_Filtering-001.png').convert('L')
filtered_img = wiener(img, (5, 5))
f, (plot1, plot2) = plt.subplots(1, 2)
plot1.imshow(img)
plot2.imshow(filtered_img)
plt.show()

Based on https://docs.scipy.org/doc/scipy/reference/generated/scipy.signal.wiener.html

Let us try to reproduce:

https://web.archive.org/web/20240320105334/https://web.stanford.edu/class/ee368/Handouts/Lectures/2014_Spring/8-Linear-Image-Processing/Wiener_Filtering.pdf#page=7

pdfimages Wiener_Filtering.pdf -png -f 7 -l 7 Wiener_Filtering

Wiener_Filtering-002.png is the original one, Wiener_Filtering-001.png the noise one and Wiener_Filtering-000.png the treated one.

Similar results as above:

from scipy.signal import wiener
import matplotlib.pyplot as plt
from PIL import Image
import warnings

img = Image.open('Wiener_Filtering-001.png').convert('L')
figure, plots = plt.subplots(1, 5)
figure.suptitle('Comparison of multiple Wiener filter size')
plot = plots[0]
plot.imshow(img)
plot.set_title('Original')
for i in range(1, 5):
    warnings.filterwarnings('error')
    try:
        filtered_img = wiener(img, i)
        hasError = False
    except:
        warnings.resetwarnings()
        filtered_img = wiener(img, i)
        hasError = True
    plot = plots[i]
    plot.set_title(f'wiener(img, {i})\n({hasError=})')
    plot.imshow(filtered_img)
plt.tight_layout()
plt.show()

The error is:

/home/benjamin/.local/lib/python3.10/site-packages/scipy/signal/_signaltools.py:1659: RuntimeWarning: divide by zero encountered in divide
  res *= (1 - noise / lVar)
/home/benjamin/.local/lib/python3.10/site-packages/scipy/signal/_signaltools.py:1659: RuntimeWarning: invalid value encountered in multiply
  res *= (1 - noise / lVar)

While https://stackoverflow.com/a/41020626 looks interesting:

from skimage import color, data, restoration
from scipy.signal import convolve2d

img = color.rgb2gray(data.astronaut())
psf = np.ones((5, 5)) / 25
img = convolve2d(img, psf, 'same')
img += 0.1 * img.std() * np.random.standard_normal(img.shape)
deconvolved_img = restoration.wiener(img, psf, 1100)
f, (plot1, plot2) = plt.subplots(1, 2)
plot1.imshow(img)
plot2.imshow(deconvolved_img)
plt.show()

with my image it does not seem to be interesting.

from skimage import color, data, restoration
from scipy.signal import convolve2d
from PIL import Image
import numpy as np
import matplotlib.pyplot as plt

img = Image.open('Wiener_Filtering-001.png').convert('L')
img = np.array(img)
psf = np.ones((5, 5)) / 25
deconvolved_img = restoration.wiener(img, psf, 1100)

f, (plot1, plot2) = plt.subplots(1, 2)
plot1.imshow(img)
plot2.imshow(deconvolved_img)
plt.show()

I am able to correctly compute RMS thanks to the Stack Overflow answer 11818358.

from PIL import Image, ImageChops
import math
import operator
import functools

def openImage(filePath):
    return Image.open(filePath).convert('L')

im1 = openImage('Wiener_Filtering-001.png')
im2 = openImage('Wiener_Filtering-002.png')

def rmsdiff(im1, im2):
    "Calculate the root-mean-square difference between two images"

    h = ImageChops.difference(im1, im2).histogram()

    # calculate rms
    return math.sqrt(functools.reduce(operator.add,
        map(lambda h, i: h*(i**2), h, range(256))
    ) / (float(im1.size[0]) * im1.size[1]))

print(rmsdiff(im1, im2))

Check initial CAI paper for details, also see the end of section IV. C. for details.

https://web.stanford.edu/class/ee368/Handouts/Lectures/Examples/8-Linear-Image-Processing/Wiener_Filtering/

It seems that thanks to #5 can deduce if denoised.

https://ipolcore.ipol.im/api/demoinfo/staticData/demoExtras/77777000278/

https://web.archive.org/web/20240320105334/https://web.stanford.edu/class/ee368/Handouts/Lectures/2014_Spring/8-Linear-Image-Processing/Wiener_Filtering.pdf#page=5 ```bash pdfimages Wiener_Filtering.pdf -png -f 5 -l 5 Wiener_Filtering ``` Unclear how to get `Wiener_Filtering-006.png` from `Wiener_Filtering-001.png`. ```py from scipy.signal import wiener import matplotlib.pyplot as plt import numpy as np from PIL import Image img = Image.open('Wiener_Filtering-001.png').convert('L') filtered_img = wiener(img, (5, 5)) f, (plot1, plot2) = plt.subplots(1, 2) plot1.imshow(img) plot2.imshow(filtered_img) plt.show() ``` Based on https://docs.scipy.org/doc/scipy/reference/generated/scipy.signal.wiener.html Let us try to reproduce: https://web.archive.org/web/20240320105334/https://web.stanford.edu/class/ee368/Handouts/Lectures/2014_Spring/8-Linear-Image-Processing/Wiener_Filtering.pdf#page=7 ```bash pdfimages Wiener_Filtering.pdf -png -f 7 -l 7 Wiener_Filtering ``` `Wiener_Filtering-002.png` is the original one, `Wiener_Filtering-001.png` the noise one and `Wiener_Filtering-000.png` the treated one. Similar results as above: ```py from scipy.signal import wiener import matplotlib.pyplot as plt from PIL import Image import warnings img = Image.open('Wiener_Filtering-001.png').convert('L') figure, plots = plt.subplots(1, 5) figure.suptitle('Comparison of multiple Wiener filter size') plot = plots[0] plot.imshow(img) plot.set_title('Original') for i in range(1, 5): warnings.filterwarnings('error') try: filtered_img = wiener(img, i) hasError = False except: warnings.resetwarnings() filtered_img = wiener(img, i) hasError = True plot = plots[i] plot.set_title(f'wiener(img, {i})\n({hasError=})') plot.imshow(filtered_img) plt.tight_layout() plt.show() ``` The error is: ``` /home/benjamin/.local/lib/python3.10/site-packages/scipy/signal/_signaltools.py:1659: RuntimeWarning: divide by zero encountered in divide res *= (1 - noise / lVar) /home/benjamin/.local/lib/python3.10/site-packages/scipy/signal/_signaltools.py:1659: RuntimeWarning: invalid value encountered in multiply res *= (1 - noise / lVar) ``` While https://stackoverflow.com/a/41020626 looks interesting: ```py from skimage import color, data, restoration from scipy.signal import convolve2d img = color.rgb2gray(data.astronaut()) psf = np.ones((5, 5)) / 25 img = convolve2d(img, psf, 'same') img += 0.1 * img.std() * np.random.standard_normal(img.shape) deconvolved_img = restoration.wiener(img, psf, 1100) f, (plot1, plot2) = plt.subplots(1, 2) plot1.imshow(img) plot2.imshow(deconvolved_img) plt.show() ``` with my image it does not seem to be interesting. ```py from skimage import color, data, restoration from scipy.signal import convolve2d from PIL import Image import numpy as np import matplotlib.pyplot as plt img = Image.open('Wiener_Filtering-001.png').convert('L') img = np.array(img) psf = np.ones((5, 5)) / 25 deconvolved_img = restoration.wiener(img, psf, 1100) f, (plot1, plot2) = plt.subplots(1, 2) plot1.imshow(img) plot2.imshow(deconvolved_img) plt.show() ``` I am able to correctly compute RMS thanks to [the Stack Overflow answer 11818358](https://stackoverflow.com/a/11818358). ```py from PIL import Image, ImageChops import math import operator import functools def openImage(filePath): return Image.open(filePath).convert('L') im1 = openImage('Wiener_Filtering-001.png') im2 = openImage('Wiener_Filtering-002.png') def rmsdiff(im1, im2): "Calculate the root-mean-square difference between two images" h = ImageChops.difference(im1, im2).histogram() # calculate rms return math.sqrt(functools.reduce(operator.add, map(lambda h, i: h*(i**2), h, range(256)) ) / (float(im1.size[0]) * im1.size[1])) print(rmsdiff(im1, im2)) ``` Check initial CAI paper for details, also see the end of section IV. C. for details. https://web.stanford.edu/class/ee368/Handouts/Lectures/Examples/8-Linear-Image-Processing/Wiener_Filtering/ It seems that thanks to #5 can deduce if denoised. https://ipolcore.ipol.im/api/demoinfo/staticData/demoExtras/77777000278/

Figure_1.png

89 KiB

Figure_1.png

156 KiB

Figure_1.png

132 KiB

Benjamin_Loison referenced this issue 2024-03-21 09:47:20 +01:00

Context-Adaptive Interpolator (CAI) #9

Benjamin_Loison added the Context-Adaptive Interpolator label 2024-03-21 10:08:01 +01:00

Benjamin_Loison referenced this issue 2024-03-21 10:08:56 +01:00

Comparing noises on images, are there clever distances? #5

Benjamin_Loison referenced this issue from a commit 2024-03-21 15:11:19 +01:00

Fix #10: Implement Wiener filter

Benjamin_Loison closed this issue 2024-03-21 15:11:19 +01:00

Benjamin_Loison commented 2024-03-21 15:15:56 +01:00

Author

Owner

Copy Link

Copy Source

With Brightness 127 and Contrast 100, get:

Before Wiener filter:

After Wiener filter:

Taking an example image region:

Before Wiener filter:

After Wiener filter:

The image looks simpler, hence is possibly denoised.

Should try applying on an image where I added Gaussian noise and measure RMS.

With `Brightness` `127` and `Contrast` `100`, get: Before Wiener filter:  After Wiener filter:  Taking an example image region: Before Wiener filter:  After Wiener filter:  The image looks simpler, hence is possibly denoised. Should try applying on an image where I added Gaussian noise and measure RMS.

image.png

8.9 MiB

image.png

6.0 MiB

image.png

62 KiB

image.png

19 KiB

Benjamin_Loison commented 2024-03-21 15:34:31 +01:00

Author

Owner

Copy Link

Copy Source

from PIL import Image, ImageChops
import math
import operator
import functools
import matplotlib.pyplot as plt
import numpy

def openImage(filePath):
    return Image.open(filePath).convert('L')

im1 = openImage('Wiener_Filtering-001.png')
im1Pixels = im1.load()
im2 = openImage('Wiener_Filtering-002.png')

def rmsdiff(im1, im2):
    "Calculate the root-mean-square difference between two images"

    h = ImageChops.difference(im1, im2).histogram()

    # calculate rms
    return math.sqrt(functools.reduce(operator.add,
        map(lambda h, i: h*(i**2), h, range(256))
    ) / (float(im1.size[0]) * im1.size[1]))

print(rmsdiff(im1, im2))

Y = []
for sigma_0 in numpy.arange(1, 20, 1):
    h_wImage = Image.new(MODE, (im1.size[0], im1.size[1]))
    h_wImagePixels = h_wImage.load()

    for i in tqdm(range(im1.size[0])):
        for j in range(im1.size[1]):
            h_wImagePixels[i, j] = round(h_w(im1, im1Pixels, i, j))

    rmsdiffValue = rmsdiff(h_wImage, im2)
    Y += [rmsdiffValue]

plt.plot(Y)
plt.show()

Unable to reduce RMS results of denoised https://web.archive.org/web/20240320105334/https://web.stanford.edu/class/ee368/Handouts/Lectures/2014_Spring/8-Linear-Image-Processing/Wiener_Filtering.pdf#page=7 maybe should try with other Q values but it would need adapting the code not to have out of bound errors.

```py from PIL import Image, ImageChops import math import operator import functools import matplotlib.pyplot as plt import numpy def openImage(filePath): return Image.open(filePath).convert('L') im1 = openImage('Wiener_Filtering-001.png') im1Pixels = im1.load() im2 = openImage('Wiener_Filtering-002.png') def rmsdiff(im1, im2): "Calculate the root-mean-square difference between two images" h = ImageChops.difference(im1, im2).histogram() # calculate rms return math.sqrt(functools.reduce(operator.add, map(lambda h, i: h*(i**2), h, range(256)) ) / (float(im1.size[0]) * im1.size[1])) print(rmsdiff(im1, im2)) Y = [] for sigma_0 in numpy.arange(1, 20, 1): h_wImage = Image.new(MODE, (im1.size[0], im1.size[1])) h_wImagePixels = h_wImage.load() for i in tqdm(range(im1.size[0])): for j in range(im1.size[1]): h_wImagePixels[i, j] = round(h_w(im1, im1Pixels, i, j)) rmsdiffValue = rmsdiff(h_wImage, im2) Y += [rmsdiffValue] plt.plot(Y) plt.show() ```  Unable to reduce RMS results of denoised https://web.archive.org/web/20240320105334/https://web.stanford.edu/class/ee368/Handouts/Lectures/2014_Spring/8-Linear-Image-Processing/Wiener_Filtering.pdf#page=7 maybe should try with other `Q` values but it would need adapting the code not to have out of bound errors.

Figure_1.svg

13 KiB

Benjamin_Loison referenced this issue 2024-03-21 17:20:13 +01:00

Inability to reproduce IPOL 77777000278 demo #16

Benjamin_Loison commented 2024-05-29 16:29:43 +02:00

Author

Owner

Copy Link

Copy Source

Related to Benjamin_Loison/PRNU_extraction/issues/1.

Related to [Benjamin_Loison/PRNU_extraction/issues/1](https://codeberg.org/Benjamin_Loison/PRNU_extraction/issues/1).

Sign in to join this conversation.

No Branch/Tag Specified

Branches Tags

master

prnu_written_as_such

evaluation

Labels

Clear labels

bug

Something is not working

Context-Adaptive Interpolator

duplicate

This issue or pull request already exists

enhancement

New feature

epic

A task that is going to take more than a day to complete.

help wanted

Need some help

high priority

invalid

Something is wrong

left for future work

low priority

medium

A task that should take less than a day to complete.

medium priority

meta

question

More information is needed

quick

A task that should take less than two hours to complete.

wontfix

This won't be fixed

No labels Context-Adaptive Interpolator

Milestone

No items

No Milestone

Projects

Clear projects

No projects

Assignees

Clear assignees

Benjamin_Loison (Benjamin Loison)

No Assignees

1 Participants

Notifications

Subscribe

Due Date

No due date set.

Dependencies

No dependencies set.

Reference: Benjamin_Loison/Robust_image_source_identification_on_modern_smartphones#10