[Wikimedia-l] AI Image Restoration

classic Classic list List threaded Threaded
3 messages Options
Reply | Threaded
Open this post in threaded view
|

[Wikimedia-l] AI Image Restoration

John Erling Blad
A quite common problem at all Wikimedia sites; we have a photo but the
quality is poor. An example is the old photo from the cabins at
Mørkedalen where a group of fighters hid out during the invasion of
Norway.[1]

I've been using some manual tools to restore images, but it is very
slow and the result is often not very good. What if we could automate
the process? The work "Deep Image Prior"[2] could be a solution, it is
quite awesome, but it is a bit too difficult to configure for most
users.

There are several such automated tools, or algorithms that can be
turned into automated tools, but the question is; do we want an
automated tool set, or do we want to keep on doing things manually?

[1] https://no.wikipedia.org/wiki/Skil%C3%B8perpatruljen_i_M%C3%B8rkedalen
[2] https://dmitryulyanov.github.io/deep_image_prior

_______________________________________________
Wikimedia-l mailing list, guidelines at: https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and https://meta.wikimedia.org/wiki/Wikimedia-l
New messages to: [hidden email]
Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, <mailto:[hidden email]?subject=unsubscribe>
Reply | Threaded
Open this post in threaded view
|

Re: [Wikimedia-l] AI Image Restoration

Fæ
For Wikimedia Commons the issue of how to handle digital restorations
has been discussed several times, there are no single solutions. The
general position is that original images, such as very old glass plate
prints from archives, should be preserved and displayed in their
original state, but there is no harm in making available restorations
as separate files that fairly and accurately represent the photograph
and remove damage or flaws. The reuser or reader should never be in
doubt that they are looking at a restoration rather than the original,
and we have to keep in mind that we do not always know whether the
reader's monitor, display device or printer may also adapt the
saturation and brightness of the image.

Though it is possible to do large batch changes for light levels,
sharpening, colour correction and so on, these remain controversial if
the original file is overwritten. This is especially true when an
artwork has been photographed or scanned.

Among recent on-wiki controversy is colorization of photographs,
especially old black and white photographs. These may be interesting,
and at times useful for illustration, but again the reader should be
in no doubt that they are not looking at "reality" and as good
practice it should always be easy for the reader to navigate to the
unenhanced original.

Fae
--
[hidden email] https://commons.wikimedia.org/wiki/User:Fae

On Tue, 28 May 2019 at 02:23, John Erling Blad <[hidden email]> wrote:

>
> A quite common problem at all Wikimedia sites; we have a photo but the
> quality is poor. An example is the old photo from the cabins at
> Mørkedalen where a group of fighters hid out during the invasion of
> Norway.[1]
>
> I've been using some manual tools to restore images, but it is very
> slow and the result is often not very good. What if we could automate
> the process? The work "Deep Image Prior"[2] could be a solution, it is
> quite awesome, but it is a bit too difficult to configure for most
> users.
>
> There are several such automated tools, or algorithms that can be
> turned into automated tools, but the question is; do we want an
> automated tool set, or do we want to keep on doing things manually?
>
> [1] https://no.wikipedia.org/wiki/Skil%C3%B8perpatruljen_i_M%C3%B8rkedalen
> [2] https://dmitryulyanov.github.io/deep_image_prior
>

_______________________________________________
Wikimedia-l mailing list, guidelines at: https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and https://meta.wikimedia.org/wiki/Wikimedia-l
New messages to: [hidden email]
Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, <mailto:[hidden email]?subject=unsubscribe>
Reply | Threaded
Open this post in threaded view
|

Re: [Wikimedia-l] AI Image Restoration

John Erling Blad
I've been playing with an idea that a a kind of batch script could be
applied to an image, and then the result of the processing becomes the
image to display, but the real image is not changed at all. Someting
like magick-script in a tag-function.[1] I have not checked how secure
it would be, but MediaWiki already use the ImageMagick package.

[1] https://imagemagick.org/script/magick-script.php


On Tue, May 28, 2019 at 10:40 AM Fæ <[hidden email]> wrote:

>
> For Wikimedia Commons the issue of how to handle digital restorations
> has been discussed several times, there are no single solutions. The
> general position is that original images, such as very old glass plate
> prints from archives, should be preserved and displayed in their
> original state, but there is no harm in making available restorations
> as separate files that fairly and accurately represent the photograph
> and remove damage or flaws. The reuser or reader should never be in
> doubt that they are looking at a restoration rather than the original,
> and we have to keep in mind that we do not always know whether the
> reader's monitor, display device or printer may also adapt the
> saturation and brightness of the image.
>
> Though it is possible to do large batch changes for light levels,
> sharpening, colour correction and so on, these remain controversial if
> the original file is overwritten. This is especially true when an
> artwork has been photographed or scanned.
>
> Among recent on-wiki controversy is colorization of photographs,
> especially old black and white photographs. These may be interesting,
> and at times useful for illustration, but again the reader should be
> in no doubt that they are not looking at "reality" and as good
> practice it should always be easy for the reader to navigate to the
> unenhanced original.
>
> Fae
> --
> [hidden email] https://commons.wikimedia.org/wiki/User:Fae
>
> On Tue, 28 May 2019 at 02:23, John Erling Blad <[hidden email]> wrote:
> >
> > A quite common problem at all Wikimedia sites; we have a photo but the
> > quality is poor. An example is the old photo from the cabins at
> > Mørkedalen where a group of fighters hid out during the invasion of
> > Norway.[1]
> >
> > I've been using some manual tools to restore images, but it is very
> > slow and the result is often not very good. What if we could automate
> > the process? The work "Deep Image Prior"[2] could be a solution, it is
> > quite awesome, but it is a bit too difficult to configure for most
> > users.
> >
> > There are several such automated tools, or algorithms that can be
> > turned into automated tools, but the question is; do we want an
> > automated tool set, or do we want to keep on doing things manually?
> >
> > [1] https://no.wikipedia.org/wiki/Skil%C3%B8perpatruljen_i_M%C3%B8rkedalen
> > [2] https://dmitryulyanov.github.io/deep_image_prior
> >
>
> _______________________________________________
> Wikimedia-l mailing list, guidelines at: https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and https://meta.wikimedia.org/wiki/Wikimedia-l
> New messages to: [hidden email]
> Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, <mailto:[hidden email]?subject=unsubscribe>

_______________________________________________
Wikimedia-l mailing list, guidelines at: https://meta.wikimedia.org/wiki/Mailing_lists/Guidelines and https://meta.wikimedia.org/wiki/Wikimedia-l
New messages to: [hidden email]
Unsubscribe: https://lists.wikimedia.org/mailman/listinfo/wikimedia-l, <mailto:[hidden email]?subject=unsubscribe>