PCD:
Actually it's a lot more complex than that.
It is true that it can adjust the hue to make the final image look better, but that's considered cheating, and doesn't look as cool as doing it the right way.
What it's really doing is it tries to find images that most closely match the actual pixels in the area in question.
In other words, if a picture is mostly green with a tan dog in the lower right corner, then it would be more likely to use that image on a section of the image that needs green in the top left of the image and tan in the lower right.
The problem is that they didn't just patent the specific algorithm they used, because one can work around an algorithm, they patented the look and feel as well, so any algorithm that produces an output that looks like that would be covered if in fact said patent is valid.
Actually it's a lot more complex than that.
It is true that it can adjust the hue to make the final image look better, but that's considered cheating, and doesn't look as cool as doing it the right way.
What it's really doing is it tries to find images that most closely match the actual pixels in the area in question.
In other words, if a picture is mostly green with a tan dog in the lower right corner, then it would be more likely to use that image on a section of the image that needs green in the top left of the image and tan in the lower right.
The problem is that they didn't just patent the specific algorithm they used, because one can work around an algorithm, they patented the look and feel as well, so any algorithm that produces an output that looks like that would be covered if in fact said patent is valid.