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

Given device split into two groups and compare both estimated PRNUs and how they evolved when consider more and more images to estimate them within each group #31

New Issue
Open
opened 2024-04-02 14:12:03 +02:00 by Benjamin_Loison · 3 comments
Benjamin_Loison commented 2024-04-02 14:12:03 +02:00
Owner
Copy Link

Consider uniform images (example sky) with RAW images.

Just verifying the data loading as it is a different file type:

image

We notice the bottom-right shadow as in GIMP.

Figure_1

Figure_2

Figure_3

2m16s to execute above algorithm.

The only interesting image examples to render are:

  • an actual image
  • a noise extraction of such actual image
  • the mean of such noise for 5 images and 50 images
  • the difference (same color scale as will be quite identical because of the expected randomness of the PRNU?) and RMS for such means (1, 5 and 50) between both subgroups?
numberOfImages=1 round(rms, 4)=0.3788
numberOfImages=5 round(rms, 4)=0.1715
numberOfImages=50 round(rms, 4)=0.0541

so the test gives the expected wanted results.

Consider uniform images (example sky) with RAW images. Just verifying the data loading as it is a different file type: ![image](/attachments/fad539a5-1bd8-494d-a9ef-80fa546872cf) We notice the bottom-right shadow as in GIMP. ![Figure_1](/attachments/0d041c5b-64af-40b7-a753-c76f52c5f91b) ![Figure_2](/attachments/31234597-5e08-479e-ab53-1e70a91fd146) ![Figure_3](/attachments/a3e1100d-646a-481b-ad60-4afa6577db16) 2m16s to execute above algorithm. The only interesting image examples to render are: - an actual image - a noise extraction of such actual image - the mean of such noise for 5 images and 50 images - the difference (same color scale as will be quite identical because of the expected randomness of the PRNU?) and RMS for such means (1, 5 and 50) between both subgroups? ``` numberOfImages=1 round(rms, 4)=0.3788 numberOfImages=5 round(rms, 4)=0.1715 numberOfImages=50 round(rms, 4)=0.0541 ``` so the test gives the expected wanted results.
image.png
1.7 MiB
Figure_1.svg
94 KiB
Figure_2.svg
920 KiB
Figure_3.svg
123 KiB
Benjamin_Loison added the
high priority
epic
 labels 2024-04-02 14:13:22 +02:00
Benjamin_Loison commented 2024-04-02 23:58:57 +02:00
Author
Owner
Copy Link

Figure_1

![Figure_1](/attachments/d32cbcbf-6226-4ef5-8cad-038e37a8ba4a)
Figure_1.svg
32 KiB
Benjamin_Loison commented 2024-04-03 15:13:51 +02:00
Author
Owner
Copy Link

Should print a 2D figure with the evolving PRNU (with all images for instance) with for each x value a representation of the distribution of PRNU pixel values to see if the PRNU does not converge to 0 globally. Pay attention to compare with the images scale.

Should print a 2D figure with the evolving PRNU (with all images for instance) with for each x value a representation of the distribution of PRNU pixel values to see if the PRNU does not converge to 0 globally. Pay attention to compare with the images scale.
Benjamin_Loison commented 2024-04-06 00:22:53 +02:00
Author
Owner
Copy Link

Methods enumerated in #25 (comment).

tv_chambolle: 3:35:52, 129.53s/it
wavelet: 06:58, 4.19s/it

np_array

np_array_0

We notice the same tiles as previously.

bilateral: 51:29:12, 1853.52s/it is insanely slow but insanely more precise it seems:

np_array

np_array_0

circles are now very clear as well as we can notice a whole image circle maybe being somehow the lens and we notice corners having issue maybe due to an optic issue. More precise description in minutes.

image

Brightness and contrast is not as good as in GIMP, see Benjamin_Loison/gimp/issues/24.

Squares are about 295x295 pixels.

Methods enumerated in https://gitea.lemnoslife.com/Benjamin_Loison/Robust_image_source_identification_on_modern_smartphones/issues/25#issuecomment-1476. `tv_chambolle`: 3:35:52, 129.53s/it `wavelet`: 06:58, 4.19s/it ![np_array](/attachments/49044015-8676-4507-8b9a-d3c3010c2abf) ![np_array_0](/attachments/7cbb3092-9f44-4027-bc2f-c1181fd42613) We notice the same tiles as previously. `bilateral`: 51:29:12, 1853.52s/it is insanely slow but insanely more precise it seems: ![np_array](/attachments/21ac3f70-75fc-452c-b79f-01014cae2b42) ![np_array_0](/attachments/fd963d85-ff00-4749-8c5a-2d92fb632fef) circles are now very clear as well as we can notice a whole image circle maybe being somehow the lens and we notice corners having issue maybe due to an optic issue. More precise description in minutes. ![image](/attachments/b613b6dd-fdb3-4284-96c2-30a64230adfb) Brightness and contrast is not as good as in GIMP, see [Benjamin_Loison/gimp/issues/24](https://gitlab.gnome.org/Benjamin_Loison/gimp/-/issues/24). Squares are about 295x295 pixels.
np_array.png
14 MiB
np_array_0.png
683 KiB
np_array_0.png
2.7 MiB
image.png
2.4 MiB
np_array.png
15 MiB
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
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 Label
epic
high priority
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: Benjamin_Loison/Robust_image_source_identification_on_modern_smartphones#31
No description provided.
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?