Inpainting, seam carving, & image retargeting are basically methods that software can use to analyze & then alter an image. They've been around a while now [almost 10 years]. For removal of objects or damaged areas in an image, apps like TeoreX Inpaint -- or similar features in an image editing app -- can be faster & more accurate than using the clone tools in an image editor.

Two weaknesses are that complex image backgrounds can confuse the software, and you're essentially asking it to create something out of nothing. MultiView-Inpaint gets around those limitations by using a method similar to color keying or green screening.

When actors appear in scene in a TV program or movie that would be impossible to physically create, e.g. in space, they film the actors in front of a completely green background, & later software makes everything green transparent, so that a computer generated background can show through when the layer with the actors on it is placed on top.

Software can also detect objects in images, doing the same sort of thing without using a green background, e.g. the software used with some Logitech webcams, usually so the moderator/gamer appears without a background overlaid on streaming in-game video on a service like Twitch or YouTube. MultiView-Inpaint does the same sort of thing only in reverse...

You import more than one image into MultiView-Inpaint, and it lets you place select areas of the background on top of objects in a composite image, effectively removing that object or objects, but without any of the guesswork you'd get with something like Inpaint. You could do the same sort of thing yourself in many image editing apps, but MultiView-Inpaint uses some of the code from TeoreX panorama software to match things up, so every shot doesn't have to have the exact same target area using a tripod.

For MultiView-Inpaint to work however you need an image that shows the background behind the object(s) you want to remove. If something is moving in front of the target you want to take a picture of, a second photo where that person for example has moved several feet should work great -- MultiView-Inpaint lets you take the clear background in one place in one shot & paste it on top of the person in the other photo. If the object you want to get rid of never moves, you *might* be able to get another shot from a different angle & use that, or you might have to use regular Inpaint.

mike  –  8 years ago  –  このコメントは役に立ちましたか？ はい | いいえ (+18)

翻訳

Changelog: http://www.theinpaint.com/download.html#multi-view-inpaint-changelog

(Unchanged since August 2014. Suppose there has been no need for any improvements.)

XP SP3
Installs. registers just fine.

Load program
Load a picture
+ (sign) to add a second picture

Error: Images must have the same pixel format.

Whatever that means?

Then I crash.

Restart, pick two more images.
Error. Crash. Restart.

Restart, pick two more images.
Error. Crash. Restart.

Finally, I pick two images that it actually doesn't complain about.
(And with that I'm a bit clueless what to do, how to do... suppose I'll have to review their website - but why should I have to, to do even the most basic operation ... ?)

Anyhow, UI is dark, which does not work well for me.

File picker (on XP at least) does not seem to respect the 'Files of type: *.jpg *.bmp *.tiff...', instead it simply shows all files.

therube  –  8 years ago  –  このコメントは役に立ちましたか？ はい | いいえ (+12)

翻訳

I installed and run it. Before I try, does this program use also RAW picture files, like .nef of Nikon cameras? The reason to use RAW is that it contains much more details 14bit of colors, more dynamic range of light.

dudi1234  –  8 years ago  –  このコメントは役に立ちましたか？ はい | いいえ (-2)

翻訳

dudi1234,

"does this program use also RAW picture files"

No... AFAIK the way things work is that RAW files contain pretty much all of the data, *uninterpreted*, from the camera's sensor. That data usually has to changed [interpreted] into a regular image file to be useful, or to be modified &/or manipulated. If you use Lightroom for example, every mod you make to an image is recorded in a script that's played when that image gets converted into something like a jpg to use in something like P/Shop -- none of the changes you make alter the original RAW file.

So in this case you'd take a series of photos, open the RAW images in Lightroom or similar, make whatever settings, apply those same settings to each of the images, export as jpg files, & use those in MultiView-Inpaint.

dudi1234, <i>"does this program use also RAW picture files"</i> No... AFAIK the way things work is that RAW files contain pretty much all of the data, *uninterpreted*, from the camera's sensor. That data usually has to changed [interpreted] into a regular image file to be useful, or to be modified &/or manipulated. If you use Lightroom for example, every mod you make to an image is recorded in a script that's played when that image gets converted into something like a jpg to use in something like P/Shop -- none of the changes you make alter the original RAW file. So in this case you'd take a series of photos, open the RAW images in Lightroom or similar, make whatever settings, apply those same settings to each of the images, export as jpg files, & use those in MultiView-Inpaint.
Save | Cancel

mike  –  8 years ago  –  このコメントは役に立ちましたか？ はい | いいえ (+6)

翻訳