[Jerman]

My apologizes if this was not in the correct forum. But seeing as this is just concept, I felt it would most appropriately fit in the Programmer's Nook. Okay, some of you may have already seen this, but I stumbled upon the following article on Slashdot:

http://science.slashdot.org/science/.../1835256.shtml

(or for the Youtube video: http://www.youtube.com/watch?v=vIFCV2spKtg)

Two incredible things were shown here. First, images can be resized and the aspect ratio can be changed without losing image information. This would be so incredibly useful for my website, for example. Right now my "banner image" is 700x250 pixels. Ideally, I would like to change it to 900x150 pixels, for a wider aspect ratio. The problem is that it is hard to get good wide shots taken, as the banners need to be actual images. With this software and some work, I could resize the images to a different ratio without losing the elements of the images.

The second thing that impressed me was the example they gave of removing objects from an image. They removed a person on a beach within a few seconds.

I sure hope that this comes out as a gimp plugin soon, or something along those lines, as both techniques would be incredibly useful, and time saving for me.

[ghoti]

It's impressive, but it's not "without losing image information". It just loses the information (or makes it up) where there are fewer edges in the image, and thus keeps those areas with more visual information (boundaries, contrast) largely intact. For your banner, you can do the same thing by stretching those parts that are less busy. You could even do an edge detect filter in photoshop to guide you (the video basically gives you a how-to).

Somebody could probably put together a plugin relatively quickly. SIGGRAPH papers are supposed to be written in a way so that a knowledgeable grad student could implement the method.

[ShadowOfGed]

This would be incredibly useful for browsers to implement, so they could intelligently resize images when the specified dimensions are larger/smaller than the original. Unfortunately, I imagine it is computationally expensive, so it may be a while before we see it implemented/used in any kind of real-life scenario.

Photoshop doesn't count, because that's already computationally expensive.