Geek Answers Handbook

Python fractal image scaling

I am in a position where relatively low resolution images are provided (via API, higher resolution images are not available) and high resolution images .

I took a look at PIL and it just works great on everything ... Besides scaling images.

It has a general resizing algorithm :

  • Nearest neighbor
  • Bilinear
  • Bicubic
  • Smoothing

I would like to use Fractal Resizing (according to Jeff's post on the horror of coding ), but, alas, PIL does not have support for this kind of resizing.

Other Google searches do not provide alternative libraries for resizing a fractal image.

Is there such a thing or do I really need to fasten and write my own fractal resizing algorithm?

I am not an expert, but from my current point of view, this looks like a pretty steep learning curve :(

If there is no such library, maybe you have some advice where you can learn about fractal compression algorithms?

+4
source share
2 answers

There are algorithms, and you definitely won't find them in Python. To get started, you can take this article:

Daniel Glasner, Shai Bagon and Michal Irini, “Super Resolution from a Single Image,” at IEEE Computer Vision International Conference, Kyoto, Japan, 2009

It is very modern, high-tech and gives promising results. If you ever get into the python implementation, release it to the public :)

+9
source

Fractal-based image scaling is still rather unusual and has not yet settled on one accepted algorithm. I'm afraid you will not find it in standard image processing libraries yet.

It is also not always preferable to bicubic. It may have artifacts that would be undesirable for some types of images. For me, the image of Jeff looks a little strange and unnatural around the sharp edges, like the right side of the nose. Better for some values ​​is "better, of course, but I would not use it for all my images."

(This also applies to other “advanced scaling technologies,” including the more well-known and more widely implemented Lanczos / Sinc method.)

+2
source

More articles:


All Articles