Despeckle leaves pink areas

[Antonio-52]

OS: Ubuntu 20.3
Digikam ver.8.10
Gmic-QT as plug-in of Digikam
tool despeckle ver.2013-02-01

Sometimes, after run despeckle, the worked image has some pink areas.
A number of "speckles" is corrected but pink areas seems to fill some white (or almost white) zones of the image.
I tried different settings but the problem still arise.
What can I do?
Is a bug of despeckle tool?

[dtschump]

This looks like out-of-range values in the resulting image.
The Despeckle filter has been made by @garagecoder , maybe he could help ?

[garagecoder]

I'll try to reproduce the problem this week, then perhaps a fix

[Antonio-52]

Thank you. I had difficulty to find how to ask help for a specific plugin of gmic-qt!

…

-- Antonio Angelo Inviato da Proton Mail mobile

-------- Messaggio originale -------- Il 20 dic 2023, 09:01, A Kelday ha scritto:

I'll try to reproduce the problem this week, then perhaps a fix — Reply to this email directly, [view it on GitHub](#192 (comment)), or [unsubscribe](https://github.com/notifications/unsubscribe-auth/ALD5RFORRG7WZFFCUBKYTKDYKKLOHAVCNFSM6AAAAABAZ5YH3GVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQNRUGAZDMOJZG4). You are receiving this because you authored the thread.Message ID: ***@***.***>

[garagecoder]

@Antonio-52 So far, I've not been able to reproduce this. Are you able to provide an example input image which has the issue?

Mathematically, I can't see how that could happen unless there's something strange about the input first. If you have time, it would also be good to know if you see the issue with g'mic in some other program (e.g. gimp plugin, or standalone).

[Antonio-52]

Ok, I'll try.

…

-- Antonio Angelo Inviato da Proton Mail mobile

-------- Messaggio originale -------- Il 20 dic 2023, 21:01, A Kelday ha scritto:

***@***.***(https://github.com/Antonio-52) So far, I've not been able to reproduce this. Are you able to provide an example input image which has the issue? Mathematically, I can't see how that could happen unless there's something strange about the input first. If you have time, it would also be good to know if you see the issue with g'mic in some other program (e.g. gimp plugin, or standalone). — Reply to this email directly, [view it on GitHub](#192 (comment)), or [unsubscribe](https://github.com/notifications/unsubscribe-auth/ALD5RFJQB67E4F6GUUZQTTDYKM7YPAVCNFSM6AAAAABAZ5YH3GVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQNRVGA2TOOJTGQ). You are receiving this because you were mentioned.Message ID: ***@***.***>

[Antonio-52]

@Kelday After some time I tested again the despeckle filter (with default parameters), but the problem still exists. I work with Digikam 8.1 and cannot update to version 8.3 (OS problem). I attached two images that have the pink problem. - 1996-09-04 11h - Valle Ihlara; 235_v1.jpg :: FILE WORKED, WITH PINK TRACES - 1996-09-04 11h - Valle Ihlara; 235_v2.jpg :: ORIGINAL FILE Thank you! -- Antonio Angelo Inviato con l'email sicura [Proton Mail](https://proton.me/). giovedì 21 dicembre 2023 11:37, Antonio Angelo ***@***.***> ha scritto:

…

Ok, I'll try. -- Antonio Angelo Inviato da Proton Mail mobile -------- Messaggio originale -------- Il 20 dic 2023, 21:01, A Kelday < ***@***.***> ha scritto: > ***@***.***(https://github.com/Antonio-52) So far, I've not been able to reproduce this. Are you able to provide an example input image which has the issue? > > Mathematically, I can't see how that could happen unless there's something strange about the input first. If you have time, it would also be good to know if you see the issue with g'mic in some other program (e.g. gimp plugin, or standalone). > > — > Reply to this email directly, [view it on GitHub](#192 (comment)), or [unsubscribe](https://github.com/notifications/unsubscribe-auth/ALD5RFJQB67E4F6GUUZQTTDYKM7YPAVCNFSM6AAAAABAZ5YH3GVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMYTQNRVGA2TOOJTGQ). > You are receiving this because you were mentioned.Message ID: ***@***.***>

[garagecoder]

Hi @Antonio-52

Unfortunately, I don't see your attached images on here or by email!
Maybe you can paste directly here in github rather than reply by email?

What I might need to do for you is create a new/test version of the despeckle filter. That way, I can add some features without breaking it for others. Hopefully I'll be able to do that this weekend.

Thanks for the feedback

[Antonio-52]

Ok, I'll try to load images to GitHub. Thanks!

…

-- Antonio Angelo Inviato da Proton Mail Android

-------- Messaggio originale -------- 13/05/24 20:11, A Kelday ha scritto:

Hi ***@***.***(https://github.com/Antonio-52) Unfortunately, I don't see your attached images on here or by email! Maybe you can paste directly here in github rather than reply by email? What I might need to do for you is create a new/test version of the despeckle filter. That way, I can add some features without breaking it for others. Hopefully I'll be able to do that this weekend. Thanks for the feedback — Reply to this email directly, [view it on GitHub](#192 (comment)), or [unsubscribe](https://github.com/notifications/unsubscribe-auth/ALD5RFMJTY3FRB6PR37ZMDLZCD65RAVCNFSM6AAAAABAZ5YH3GVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDCMBYGQ4TQOJQGM). You are receiving this because you were mentioned.Message ID: ***@***.***>

[Antonio-52]

• _v2 is the original image
• _v1 is after working with despeckle filter

[garagecoder]

@Antonio-52 I still can't reproduce this, even with those images. I think something is going wrong your side which is probably not due to this filter. I expect you could find other filters in g'mic which would break in similar ways.

I've added a new test filter called "Despeckle [test]" which has two clipping options. Both will stop any values outside 8bit range [0,255]. ClipIn clips before the filter is applied, ClipOut happens after.

If using ClipIn solves it, I think that shows something goes wrong before the filter. If just ClipOut solves it, then the filter is going wrong in a way I don't yet understand.

You will need a very recent version of g'mic to try that test filter!