feature #47: binarizeEdgePlus

[zvezdochiot]

Hi @vigri and @lightsilverberryfox .

⚠️ This PR is just a recipe. I was only able to build binarizeEdgePlus into the general functionality. At the same time, I use the Sauvola options, which I don’t think is correct (the new threshold requires smaller area values ​​and larger coefficient values). Due to this, projects with a new threshold are safely saved. Use this PR as a recipe.

Good Luck.

---

Example

Origin	Otsu threshold	Sauvola threshold	Wolf threshold	EdgePlus threshold	EdgePlus threshold	EdgePlus threshold
(screenshot from 600dpi scan)	(default)	(default: ws:200 c:0.34)	(default)	(default: ws:200 c:0.34)	(manual: ws:20 c:0.80)	(manual: ws:5 c:0.99 Equalize:0ff Savitzky-Golay:off)

Links

• Build this PR
• OCR forum (RU): ScanTailor Advanced; ideas for improving the clarity of scans
• Scan Tailor (part 2)

Closes #47

[zvezdochiot]

Hi @vigri , @lightsilverberryfox .

The usefulness of the new thresholds has been confirmed (worst part of the scanned material):

Origin	Otsu	Sauvola	Wolf	Wolf	EdgePlus	BlurDiv
(600dpi)	(default)	(default)	(default)	(ws:20,c:0.75)	(ws:20,c:0.75)	(ws:20,c:0.75)

See threshold BlurDiv.

[zvezdochiot]

[Feature request] thresholdAdjustment in other threshold. #59

[zvezdochiot]

Mod epdbta of release 1.0.18: https://github.com/ImageProcessing-ElectronicPublications/scantailor-advanced/releases/tag/v1.0.18

[alfish2000]

Hi!
Is there any chance that the PRs here get merged and e.g. a new version 1.0.19 is released?
It gets confusing because it seems that this PR is now released here https://github.com/ImageProcessing-ElectronicPublications/scantailor-advanced as 1.0.18

[zvezdochiot]

@alfish2000 say:

Is there any chance that the PRs here get merged and e.g. a new version 1.0.19 is released?

I don't hope so. Therefore, I release mods for existing releases. ;)

[ghost]

Okay, the code seems to be fine, I don't understand how the EdgePlus algorithm works though... But if it works it should be okay. Let's see if @vigri has something to say.

[mara004]

Is there any chance that the PRs here get merged and e.g. a new version 1.0.19 is released?

I don't hope so. Therefore, I release mods for existing releases. ;)

I'm confused by your "I don't hope so."
Are you still OK with this being merged and included in a release here eventually?
Otherwise, I think you should close the PR, mark it as draft, or similar.

[zvezdochiot]

Hi @mara004 .

I don't have enough experience with Qt and XML. If there are bugs, it will be difficult for me to eliminate them. If this PR enters the mainstream, I will be happy. Doubts are caused purely by lack of skill (especially in XML).

PS: During testing, I did not notice any failures.

[zvezdochiot]

Hi @mara004 , @lightsilverberryfox , @vigri .

Demo of new thresholds:

Origin	Otsu	EdgePlus
150dpi	600dpi (d: 16)	600dpi (d:16, ws: 20, c: 0.99)

[zvezdochiot]

Hi @mara004 , @lightsilverberryfox , @vigri .

Github Action: build-linux-qt5:

...
Err:12 http://azure.archive.ubuntu.com/ubuntu jammy-updates/main amd64 libinput-bin amd64 1.20.0-1ubuntu0.2
  404  Not Found [IP: 52.154.174.208 80]
Err:13 http://azure.archive.ubuntu.com/ubuntu jammy-updates/main amd64 libinput10 amd64 1.20.0-1ubuntu0.2
  404  Not Found [IP: 52.154.174.208 80]
...
E: Failed to fetch http://azure.archive.ubuntu.com/ubuntu/pool/main/libi/libinput/libinput-bin_1.20.0-1ubuntu0.2_amd64.deb  404  Not Found [IP: 52.147.219.192 80]
E: Failed to fetch http://azure.archive.ubuntu.com/ubuntu/pool/main/libi/libinput/libinput10_1.20.0-1ubuntu0.2_amd64.deb  404  Not Found [IP: 52.147.219.192 80]
E: Unable to fetch some archives, maybe run apt-get update or try with --fix-missing?
Error: Process completed with exit code 100.

http://azure.archive.ubuntu.com/ubuntu/pool/main/libi/libinput/:

name	date	size
libinput-bin_1.20.0-1_amd64.deb	2022-02-22 23:45	19K
libinput-bin_1.20.0-1_i386.deb	2022-03-02 14:09	24K
libinput-bin_1.20.0-1ubuntu0.1_amd64.deb	2022-05-02 06:13	19K
libinput-bin_1.20.0-1ubuntu0.1_i386.deb	2022-05-02 06:13	24K
libinput-bin_1.20.0-1ubuntu0.3_amd64.deb	2023-06-16 18:24	19K
libinput-bin_1.20.0-1ubuntu0.3_i386.deb	2023-06-16 18:24	24K

---

Thu, 13 Jul 2023 15:49:44 +0300

Resolved:

• CMake / build-linux-qt5 (push) - ok
• clang-format-check / build (push) - ok

[ghost]

Okay, @zvezdochiot

After some testing and reviewing, I believe the PR is ready to be merged, I didn't encounter any issues and this could be a very useful feature, I would normally wait for @vigri to offer their review, however he hasn't been here for three months already, so, I don't know what to do.

If you believe it's ready to be merged, I'll merge it. I can tell you put a lot of effort in this PR and wanted to say thank you, you have also offered your input on other issues and bugs, so if vigri sees this, he may want to consider sending you an invitation to our organization so you have write access (I can't invite you myself, I don't have owner permissions).

Is issue #59 also solved by this PR?

[vigri]

Overall I see no problems. Please see my comments. Especially the one with the qt-version.

[vigri]

I can tell you put a lot of effort in this PR and wanted to say thank you, you have also offered your input on other issues and bugs, so if vigri sees this, he may want to consider sending you an invitation to our organization so you have write access (I can't invite you myself, I don't have owner permissions).

apologies very much.

Thank you very much @zvezdochiot.
As @lightsilverberryfox said: If you are interested, I can send you an invitation. Please let me know.

Regarding this PR: Pleasse see my "review".

[zvezdochiot]

@lightsilverberryfox say:

Is issue #59 also solved by this PR?

This PR closes the issues:

• Feature Request: Binarization Algorithms #8
• [Feature request] binarizeEdgePlus #47
• [Feature request] thresholdAdjustment in other threshold. #59

PS: Me, @plzombie and @trufanov-nok will oversee the STA fork at organization https://github.com/ImageProcessing-ElectronicPublications. Participation in https://github.com/ScanTailor-Advanced is not required.

[vigri]

Thank you very much again for this PR!

[mara004]

PS: Me, @plzombie and @trufanov-nok will oversee the STA fork at organization https://github.com/ImageProcessing-ElectronicPublications. Participation in https://github.com/ScanTailor-Advanced is not required.

Wouldn't it be better to join efforts in one repo/organization, instead of all the fragmentation?

[zvezdochiot]

@mara004 say:

Wouldn't it be better to join efforts in one repo/organization

Maybe, but not now. I myself specialize in console utilities (GUI is not mine), @plzombie is busy with depress (it is more important to him), @trufanov-nok is in charge of STU. Everyone is busy with their projects.

PS: Until I figure out the STA (XML) projects, my help is over. I won't figure it out until I see how it's done. Such are the things.

[zvezdochiot]

Hi @mara004 , @lightsilverberryfox , @vigri .

New version?
1.0.19?
I need for forums.

Current:

scantailor-advanced/version.h.in

Line 6 in b0788fc

#define VERSION "1.0.18"

[vigri]

Hi @mara004 , @lightsilverberryfox , @vigri .

New version? 1.0.19? I need for forums.

Current:

scantailor-advanced/version.h.in

Line 6 in b0788fc

#define VERSION "1.0.18"

I've uploaded 1.0.19 now.
https://github.com/ScanTailor-Advanced/scantailor-advanced/releases/tag/v1.0.19

[ireun]

Guys it would be great to have the docs improves to explain the new thresholds, maybe with examples or when to use what. :)

[zvezdochiot]

Hi all, @ireun .

⚠️ Don't clutter this repository!
Documentation can be updated, but examples and stuff only in a separate repository. In this link only!

"Recommendations from the author":

1. EdgeDiv is the most balanced combination of Otsu threshold prefilters. This is a "universal solution" when you need clearly drawn letters without thickening and narrowing. Requires additional cleaning of the result from "garbage". Initial parameters: d:0, ws:20, c:0.75.
2. EdgePlus is an ill-balanced Otsu threshold prefilter. It is used when it is necessary to highlight the weakly expressed contours of letters at any cost. Creates a lot of "garbage", requires thorough cleaning. Initial parameters: d:0, ws:5, c:0.99.
3. BlurDiv is an ill-balanced Otsu threshold prefilter. Whitens the background. It is used when it is necessary to get rid of a very noisy background and at the same time, if possible, preserve the characters. Initial parameters: d:0, ws:50, c:0.75.

-- Google Translit