Commit Graph

83 Commits

Author SHA1 Message Date
Benjamin Loison
83a12f4e0a
Add and use SKY and WALL 2024-04-17 14:26:13 +02:00
Benjamin Loison
e080e841f5
Add and use raiseNotFlatFields 2024-04-17 14:10:11 +02:00
Benjamin Loison
a6ef3977dd
Add comment concerning printing {min,max}Color 2024-04-16 16:48:29 +02:00
Benjamin Loison
8ca1972ddb
Correct raise not flat-fields processing 2024-04-16 16:48:00 +02:00
Benjamin Loison
817c016815
Give a try to get image size with rawpy but it seems far slower
As got:

```
Filtering images:   0%|▏                                                                               | 21/8156 [00:11<1:17:28,  1.75it/s]
```
2024-04-16 04:06:49 +02:00
Benjamin Loison
04e0de4d33
Try making not flat-images compatible 2024-04-16 03:59:26 +02:00
Benjamin Loison
3a4f774e28
Make denoise arguments depend on denoiser 2024-04-16 03:27:26 +02:00
Benjamin Loison
caf7025504
Correct typos in previous commit 2024-04-16 03:15:53 +02:00
Benjamin Loison
d4e13ed123
Compute automatically extreme values 2024-04-16 03:11:38 +02:00
Benjamin Loison
a8fa053687
#49: Get PRNU by using RAW images 2024-04-16 03:02:58 +02:00
Benjamin Loison
6251c487e2
Add temporarily maxGreen computation
And got:

```
Denoising images:   0%|                                                                      | 0/100 [00:00<?, ?it/s]maxGreen=2275
Denoising images:   1%|▌                                                             | 1/100 [00:00<01:11,  1.38it/s]maxGreen=2377
Denoising images:   2%|█▏                                                            | 2/100 [00:01<01:18,  1.24it/s]maxGreen=3468
Denoising images:  19%|███████████▌                                                 | 19/100 [00:11<00:46,  1.75it/s]maxGreen=4908
Denoising images: 100%|████████████████████████████████████████████████████████████| 100/100 [00:57<00:00,  1.74it/s]
```
2024-04-16 02:51:27 +02:00
Benjamin Loison
90df8a7cb9
Verify raw.raw_pattern as raw.color_desc does not seem enough
As got:

```bash
python3 extract_noise.py
```
```
Denoising images:   0%|                                                                                           | 0/8156 [00:00<?, ?it/s]/mnt/HDD0/raise/nef/ra2c888f8t.NEF
Denoising images:   0%|                                                                                           | 0/8156 [00:00<?, ?it/s]
Traceback (most recent call last):
  File "/home/benjamin/robust_image_source_identification_on_modern_smartphones/datasets/raise/extract_noise.py", line 73, in <module>
    treatImage(imageFileName)
  File "/home/benjamin/robust_image_source_identification_on_modern_smartphones/datasets/raise/extract_noise.py", line 46, in treatImage
    assert np.array_equal(raw.raw_pattern, np.array([[3, 2], [0, 1]], dtype = np.uint8))
AssertionError
```
2024-04-16 01:06:59 +02:00
Benjamin Loison
120eaf563f
WIP: Add .nef support 2024-04-15 23:39:47 +02:00
Benjamin Loison
a0b366e98a
Use shebang compatible with venv
See [Benjamin-Loison/cpython/issues/10](https://github.com/Benjamin-Loison/cpython/issues/10).
2024-04-15 23:17:04 +02:00
Benjamin Loison
bddfa4c8d5
Clean datasets/raise/extract_noise.py 2024-04-15 16:16:29 +02:00
Benjamin Loison
a09f2ba1c0
#29: Add Nikon D7000 4946x3278 PRNU estimation from NEF to PNG images 2024-04-15 16:06:05 +02:00
fed5698e84
Update annotations 2024-04-09 03:43:17 +02:00
Benjamin Loison
bc06612501
Make extract_noise.py easier to configure by providing denoiser 2024-04-08 04:52:29 +02:00
Benjamin Loison
8748e648b5
Remove no more relevant commented code for a test 2024-04-06 00:06:13 +02:00
Benjamin Loison
15c41c0463
Add and use imageDenoisedNpArray to ease using other denoisers 2024-04-06 00:03:50 +02:00
Benjamin Loison
b18080e9f5
#33: Switch to mono-thread 2024-04-06 00:00:24 +02:00
Benjamin Loison
a83e8bb1e0
Replace tabs with spaces in datasets/raise/extract_noise.py 2024-04-05 23:57:37 +02:00
606821d45e
Add show_mean_noise.py 2024-04-03 22:18:18 +02:00
b354f59ac5
Make parallel production ready 2024-04-03 19:15:21 +02:00
0f394faa61
Multicores prototype 2024-04-03 18:20:31 +02:00
faed1e33e8
#33: Add datasets/raise/extract_noise.py 2024-04-03 18:20:08 +02:00
4dd52aae90
Use plt.imsave instead of toPilImage 2024-04-03 16:57:49 +02:00
d3af12ce3e
Try to have clear figure for thesis audition 2024-04-03 15:20:24 +02:00
aafb7ebc92
Plot an exhaustive curve to make sure that results are not just by chance 2024-04-02 23:58:41 +02:00
631ed6de34
#31: Establish expected wanted results about finding an identical PRNU for both subgroups 2024-04-02 21:05:15 +02:00
f09665f856
Make first rendering with both subgroups 2024-04-02 19:58:59 +02:00
ea00f42c58
Add datasets/raise/split_and_compare_prnus_of_subgroups.py 2024-04-02 18:45:40 +02:00
90a806a7e2
Move webpage articles to datasets/raise/website/ 2024-04-02 18:31:27 +02:00
0d6a956db8
Shorten figure titles 2024-04-02 11:03:47 +02:00
f0db61efe2
Make single figure to compare multiple splitNXN values 2024-03-29 13:47:27 +01:00
81852d5dc6
Make Gaussian noise in getPrnuShownAsSuch only on text 2024-03-29 13:22:36 +01:00
ec7c5c6688
Add gaussianNoise to getPrnuShownAsSuch but applies on the whole PRNU image 2024-03-29 13:16:50 +01:00
2d96bdc225
Not correct generalized SPLIT_N_X_N due to imageWithoutPrnuNpArrayTile 2024-03-29 13:11:24 +01:00
96bbd50a3b
Generalize with a partial image SPLIT_N_X_N 2024-03-29 12:57:56 +01:00
4bf9ef8206
Add splitNxN compatible with 1 and 4 2024-03-29 12:44:42 +01:00
60eceb575e
Simplify grayscale rendering 2024-03-29 12:16:13 +01:00
b375acbb3a
Use v{min,max} for enforcing matplotlib colormap
If it proceeds linearly to *covers the complete value range of the supplied data* then doing so is unnecessary and even if it is another not very different scale transformation then it is still fine for my goal.
2024-03-29 12:11:41 +01:00
5fa61f7ff8
Make PRNU compatible with 4x4 split 2024-03-29 11:23:48 +01:00
8351e46437
Comparison of an image without and with Gaussian noise and PRNU 2024-03-29 01:55:46 +01:00
2f66e82f22
Show both RMS and RMS normalized 2024-03-29 01:47:01 +01:00
3ccec5bbd0
Add RMS computation 2024-03-29 01:35:48 +01:00
f297060f42
Split images in 4x4 to increase PRNU estimation accuracy 2024-03-29 01:21:26 +01:00
9b57d3441c
Render PRNU estimate taking into account all images 2024-03-29 01:10:31 +01:00
d59a251b1f
Revert to some extent previous commit 2024-03-29 01:07:48 +01:00
4382b3d649
Add PRNU_FACTOR 0.1 and 0.01 view on a single figure 2024-03-29 01:05:28 +01:00