On super-resolution, hqnx and improving Day of the Tentacle..

Inspired by a great demo website for a super-resolution paper, I’ve started to research super-resolution algorithms.

At work, I’m usually not doing “natural” images, mostly look at structures like texts, codes (data matrix ftw), geometric shapes,.. and if I’m not lucky, features (like one line from a barcode) are rather tiny (2-3px is my personal minimum). It is possible to work with these kinds of algorithms, but it’s hard if you are looking at contours and doing machine learning and rotating and stretching your pictures..

And what I’ve seen has fascinated me.

So I started really slow: With implementing hqnx. In LabVIEW.
It’s a lookup-algo for scaling pixel graphics, it’s not smart at all, but it’s.. pretty.

I like pretty things 🙂

Pictures scales/hqnx

The scaling is exceptionally good where the lines in the original image are very crisp.

Umm… so it was a bit boring to code, but okay. It works.

And yesterday my boss told me about SCUMMVM and Day of the Tentacle.
(On a Friday. Jesus. Does he ruin my weekends on purpose?)

Well, I was instantly hooked. And how to get closer to a piece of open source software than to modify it?
So I decided to download the source.
And a few minutes of reading the code led me to their graphic filters.

Yeah, they have hqnx. And a bunch of other algorithms that will be the beginning of yet another wikipedia session, until it turns into a mad coding session and then the weekend will be over..

So much for hacking fractal based super resolution algos.
Thanks a lot, boss.


Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: