Back into image coding

It’s been a long time since I have written any serious imaging code — basically since I started my current job I simply haven’t had the time, and have found other coding persuits to be more interesting. Hence the handover of Panda to Evan.

This last weekend though I had a good reason to start caring about images again. The experimental ov519 driver for Linux which is out there is a bit odd because the driver delivers JPEG compressed data through the /dev/video[0-9]+ device file. This is because that’s what the underlying hardware does. This doesn’t match the requirements of the published video4linux interface however.

I’ve therefore whipped up a minimal video4linux application which twiddles the ioctls, reads the JPEG compressed data, and then decompresses it. Well, that’s the theory at least — the in-memory decompression is this afternoon’s task, and I haven’t finished it yet.

The next step after that is a simple kernel module which takes the decompressed data from user space, and offers it back up via a video4linux device file. That means that all the video4linux applications which you have grown to love will continue to work, new applications can be written that know about JPEG, and the decompression doesn’t have to be done in kernel. It appears that the dude who wrote motion has a simple kernel module which I can leverage as well.

Very cool, and it’s nice to make some visible progress for once.

On the work front, we’ve stopped design on my latest project, and started coding. It feels really good to actually get to write some code for once, and my moral is improving. The first milestone was met today, basically on time…

On the organizations front, LCA 2005 seems to be progressing well (given that we’re just less than a year out from the conference now), there’s lots happening with Linux Australia regarding the Free Trade Agreement with the US and other things, and AUUG is progressing with getting their annual conference under way. I need to chase some AUUG issues up regarding contacting Australian computer science departments to let them know about the conference, but I’ll be looking into that this evening.