stillhq.com : Mikal, a geek from Canberra living in Silicon Valley (no blather posts)

Do you think there is something wrong with my FFT?

Tue, 26 Jul 2005 01:09:00 GMT

I'm no genius, but I don't think FFTs are meant to look like this:

Tags for this post: phd()

Comment

Bingo!

Tue, 28 Jun 2005 00:35:00 GMT

So, the code for YUV420 to RGB24 conversion now works, which is good. It's a little embarrassing that this took a couple of hours. At one point I mistook y for Y, and had a sign error. Oh well...

Technorati tags for this post: phd opensource linux video4linux yuv rgb convert color space

Comment

Ok, I've finally snapped

Mon, 27 Jun 2005 17:46:00 GMT

I was just writing some code to manipulate a bitmap, something I need to do for my PhD (specifically, I'm converting YUV420P to RGB24). Anyhoo, the image isn't looking right, so before I go to sleep on it I thought I would dump the YUV values next to the RGB ones and see if I could eyeball the problem.

I swear that as I was writing this I was thinking -- "I'll let the compiler optimise and write for readability". Do you think I succeeded?

for(x = 0; x < 320; x++)
    {
      for(y = 0; y < 240; y++)
	{
	  raster[((y + 480) * 640 * 2 + x + 640) * 3] = 
	    realfile[640 * 480 + (y * 320 + x)];
	  raster[((y + 480) * 640 * 2 + x + 640) * 3 + 1] = 
	    realfile[640 * 480 + (y * 320 + x)];
	  raster[((y + 480) * 640 * 2 + x + 640) * 3 + 2] = 
	    realfile[640 * 480 + (y * 320 + x)];

	  raster[((y + 480) * 640 * 2 + x + 640 + 320) * 3] = 
	    realfile[640 * 480 + (320 * 240) + (y * 320 + x)];
	  raster[((y + 480) * 640 * 2 + x + 640 + 320) * 3 + 1] = 
	    realfile[640 * 480 + (320 * 240) + (y * 320 + x)];
	  raster[((y + 480) * 640 * 2 + x + 640 + 320) * 3 + 2] = 
	    realfile[640 * 480 + (320 * 240) + (y * 320 + x)];
	}
    }

Here's the image I get at the moment:

See how the RGB image on the left isn't colorful enough? That pen there is lime green. Anyways, if anyone can see the problem and wants to give me a hint, that would be cool too.

Update: It would seem that the terms YUV420, YUV420P, iYUV, and I420 are all interchangeable, to add to the fun.

Technorati tags for this post: phd yuv rgb color space convert code

Comment

Huffman coding

Tue, 07 Jun 2005 00:47:00 GMT

Some links on Huffman coding:

Technorati tags for this post: phd huffman coding theory link

Comment

Random learning for the day

Tue, 07 Jun 2005 00:06:00 GMT

So, I didn't realise that Group 3 Fax Compression, probably the most common compression applied to those little bits of paper a lot of people transmit around the planet, is just run length compression. I always assumed it was better than that, especially as the libtiff implementation differentiates between the two.

No matter where you go, there you are.

Technorati tags for this post: phd compression group 3 fax ccitt tiff run length

Comment

The JPEG still picture compression standard

Thu, 02 Jun 2005 02:11:00 GMT

PhD night this week is brought to you by an understanding wife and Cite-seer...

Original paper discussing the JPEG standard and it's standardisation progress. Citations. Interesting looking citations include:

I can't provide a better link than the ACM redirect link, as the ACM portal seems to be suffering at the moment:

Error Diagnostic Information

An error occurred while attempting to establish a connection to the service.

The most likely cause of this problem is that the service is not currently running. 
You can use the 'Services' Control Panel to verify that the service is running and 
to restart it if necessary.

Windows NT error number 2 occurred.

Technorati tags for this post: phd acm jpeg compression image

Comment

My thesis worries are over

Wed, 25 May 2005 16:54:00 GMT

Google has once again saved the day:

I can buy a thesis from at least five places, including the quite reputable Ebay!

Technorati tags for this post: phd thesis online purchase ebay google

Comment

First to publish, academic articles, and blogging

Mon, 23 May 2005 18:18:00 GMT

So, I've been wondering about this for a little while. Where do I draw the line between blogging, and putting in jeopardy the first publication date for my academic papers? My understanding is that my thesis needs to be unique and original at the time of publication, but that if I blog bits of it before then, then I put that at risk because someone else might write a paper about it and publish it before me. Then again, how does blogging differ from progress seminars?

Discuss.

Technorati tags for this post: phd publication blog

Comment

Please be gentle

Mon, 23 May 2005 00:18:00 GMT

I just submitted the first draft of my first academic paper to my PhD supervision team. I hope their gentle...

Technorati tags for this post: phd article

Comment

PhD night

Thu, 19 May 2005 02:28:00 GMT

My loving and very patient wife Catherine decided after the conference was finished that I should continue to use the organising night (Thursdays) to hang around university, but this time for my PhD. I liked the idea, so here I am on my first night in the office at ANU plodding away. I'm actually being fairly productive, which is nice.

For my PhD I agreed to do 20 hours of study a week. To be honest I do more some weeks, and less others, but I'm pretty sure without having tracked it at all that I'm doing more than the 20 hours. This evening will just mean that I can spend some more time with the kids on the weekends, which will be nice.

I suspect that one of the reasons it doesn't feel like I'm progressing a lot at the moment is that I am in a writing up stage, and I'm working at approximately half the rate of a normal PhD student. I'm a little over a year in, which means that I've done the equivalent of six months for a full time PhD student. I wonder how much they've got done within their first six months?

Technorati tags for this post: phd progress

Comment

Fast Fourier transforms for gnuplot

Thu, 12 May 2005 00:12:00 GMT

Does anyone know of a good way of doing Fast Fourier Transforms with gnuplot? I can't find anything online which implies that it's possible, so I'm feeling a bit sad at the moment.

Technorati tags for this post: phd fast fourier transform fft gnuplot

Comment

Argh! CVS!

Wed, 11 May 2005 00:45:00 GMT

So, after fighting CVS for the afternoon I now have all the data back into a consistent state (most of the data is generated on my research machine in my office at university, whilst the editing of reports is done on my laptop where it happens to be). Now I'm seriously considering changing from CVS as it was a total pain... Then again, there are a lot of tools that I have hanging off CVS now, so perhaps I'll wait just a little bit longer.

Technorati tags for this post: phd cvs

Comment

Thunk

Tue, 10 May 2005 16:51:00 GMT

That's the sound of me dropping in new data for my PhD report that I am working on. I discovered that a bunch of my colour data wasn't being generated properly, which put a bunch of my conclusions into doubt. I've spent a bunch of time regenerating gigabytes of images so that I can rerun my analysis scripts, and now I just need to make sure that I am still making conclusions which are justified. I'm hoping to have a first draft of some form to my PhD supervision team by the end of the week. Famous last words.

Technorati tags for this post: phd

Comment

Getting back into it

Wed, 04 May 2005 23:54:00 GMT

Realistically I lost about four months of this year on my PhD to running linux.conf.au. Now I need to get back into things. I currently in the process of writing up my first academic paper, and need to make sure that my results, and more importantly the analysis of those results is correct. So, back so staring at a monitor I suppose.

Technorati tags for this post: phd

Comment

More academic paper progress

Mon, 21 Mar 2005 01:38:00 GMT

Well, I've been a little too slack recently on working on my first academic paper for my PhD thesis. This is mainly because I've been distracted helping run linux.conf.au 2005, but it's also been because I've been grappling with issues of style for an academic paper, of which I have not written many in the past. One of my supervisors, Tridge had an interesting suggestion -- just go with plain English. It's much easier to read, and coveys the information better as well.

Anyways, I was strong and sat down for three or four hours this afternoon and got some serious writing in. It helped a lot, as I think I've finally broken the writers block which was holding me up. Ironically for a computer science paper, I wrote it on a piece of dead tree. There's something daunting about a big pad of blank paper which makes it much more motivational to at least get something down -- even if it's wrong. You can always change it later.

Don't get me wrote though. I had actually written some stuff before, about eight or so formatted pages. It's just than in one day I think I'll have come close to doubling that length, and got much closer to a completed document.

Why the first paper of my thesis? Well, I'm going for the American style of thesis, where you take a series of papers you've written during your studies and got published in peer reviewed journals, and you staple them together to produce a thesis. This is a lot safer than the more traditional style of big bang thesis, as the uniqueness test occurs at the time of individual paper publication, not later when the whole thesis comes out. This mitigates a lot of the risk of spending a big chunk of your life producing something which isn't unique enough to count.

Technorati tags for this post: phd academic paper author

Comment

SnoopyPro vs Sniff 2000

Mon, 13 Dec 2004 02:26:00 GMT

I've been playing with SnoopyPro again, after having put it on the shelf for about six months. Anyways, I needed to compile it again to make usblogdump work with some URBs I hadn't implemented the first time around. I can't get it to compile with the latest version of the Windows Device Driver toolkit. Now I'm playing with Sniff 2000, which seems to be the next iteration of the code base. I'll let you know how things go...

Comment

Started writing up my compression stuff

Mon, 13 Dec 2004 02:24:00 GMT

Well, I finally started writing up my compression stuff the other day. It's a little bit exciting, as there are a bunch of tools to learn on the way. I've conquered gnu plot, resizing eps (try hacking the postscript with a sed script is my advice), and now I'm onto an extended battle with latex. My newfound ACM membership helps here though -- there are a bunch of latex tutorial articles from Linux Journal in the ACM archive.

Comment