h a l f b a k e r yNaturally low in facts.
add, search, annotate, link, view, overview, recent, by name, random
news, help, about, links, report a problem
browse anonymously,
or get an account
and write.
register,
|
|
|
The National Archive has three separate images of Lincoln
printed from the same negative: (see link)
Each of the prints has age damage, but
the majority of the damage is in different areas on each
print.
There should be software that uses the alignment
mechanism of focus stacking, pseudo
HDR, superresolution
etc to pick the common areas from 2 out of 3 or more
aligned images, producing a composite that has all the
blemishes automatically removed. (aka a '2 out of 3
consensus function')
If this exists, I can't find it. I expect a lot of 'this is already
baked' followups but I've spent a day looking for this and
not found anything that is exactly this product - just things
that could be made to do something similar with some
manual effort.
Note I am not talking about any of the other restoration
techniques such as in-painting a marked area with data
from other locations in the image - this is a single-function
image merge program I'm suggesting, which only restores
original data from multiple original prints. Other cleanup
techniques could be applied later once as much of the
original data is recovered as possible.
Images of Lincoln
https://www.flickr..../72157624246586102/ Various contemporary prints that could be restored [gtoal, Jul 25 2017]
Tourist removal as suggested by [MB]
http://lifehacker.c...hop-scri-1245505649 I literally did a Google search for [photoshop remove tourists] (after posting my anno, because I was that sure) and clicked the first result. Note that it is automatic and it suggests using a dozen photos. For this application I would suggest trying mode instead of median. [notexactly, Aug 15 2017]
[link]
|
|
Hmm. This would be easy to do in Photoshop in at least two ways: |
|
|
(1) Put the images into separate layers and erase flaws in the topmost layer. |
|
|
(2) There's also a method (I forget the details) whereby you can take multiple shots of a scene with moving things in it (eg, a tourist hot-spot), and then remove objects that differ between shots (ie, filter out the tourists to leave only the scenery). |
|
|
Yep - I've used (2) but they way they work is on TWO images
and you have to manually paint over the area you want
removed from one and replaced by the other. By using 3
images you don't need that manual step... Standard
panorama-style blending *ought* to do this fairly easily, it's
just that no-one appears to have built a custom app for this
particular use-case of the algorithm. |
|
|
// I've used (2) but they way they work is on TWO images// |
|
|
No, I've used it with multiple images. Something to do with a stack. If you Google "Photoshop remove tourists" or something like that, you'll find it. |
|
|
Yep, I'm familiar with that (I use photoshop) - but it's by
no means automatic. Also any solution like that does not
take care of blending the image - for example if one of
the source images is a little darker than the other,
pano/hdr style software will adjust the intensity of the
patch taken from one image before inserting it in the
other. And yes, you can do this with multiple images, but
they're just being used as more sources of data, they're
not being used to determine what pixels are common to
more than one source. As I said, there are a lot of
solutions that are in the right ballpark, but none (that I
know of yet) that are quite exactly what I've described.
(btw I also own a program called "Multiview Inpaint"
which comes very close and automates most of the
'tourist removal' trick that can be done manually in
Photoshop, but it is still not the fully automatic repair I'm
suggesting. This may just be too much of a niche
function to have been worth anyone's time to write... |
|
|
" This may just be too much of a niche function to have been worth anyone's time to write... " |
|
|
Yeah .... Almost like the Halfbakery ;-) |
|
|
No, it is automatic, and yes, it does take the consensus
from multiple images. See [link]. |
|
| |