canspice.org

For more coverage see Alasdair’s blog.

Today was Day One of ADASS XIV in Pasadena, California. It kicked off at 8:30 with a plenary speech by George Helon about IPAC. Next up was a talk about blind deconvolution techniques in image processing by Tony Chan. Deconvolution techniques are used to try to recover data by removing blurring caused by the actual act of looking at something. If you use Photoshop and have ever done a Gaussian blur, what you’ve done is taken a Gaussian function (looks like a bell-curve, but in two dimensions) and convolved it with your original image, causing a blurred image. Deconvolution techniques go backwards — you have a blurred image and some sort of point-spread function (for a Guassian blur the PSF is the Gaussian function), do some math and get a sharpened image out the end. This technique is considered “blind” because you don’t know what the point-spread function is — most deconvolution techniques force you to make a guess as to what the PSF is. Chan’s technique uses what’s called Total Variation. In any case, he showed some examples and they looked pretty impressive, getting out sharpened images that look nearly as good as the case where you know what the PSF is. He came at it from a non-astronomy point-of-view and admitted that it would be tough work for astronomy where the PSF isn’t a simple function like a Gaussian. It would also be tough because sometimes in astronomy the PSF isn’t constant over an image… I don’t think that would be much of a problem because the PSF doesn’t vary that much. Well it does, but deconvolution would probably still be better than nothing, I say naively.

Next up was visual data mining from Julio Valdes at NRC Canada. This one essentially was a talk about mapping your data set into a set of real numbers, then graphing that set of numbers in some kind of 3D fashion. He said that they used virtual reality somehow which sounded kind of Gibsonesque, but it must have translated horribly for the presentation. I have written in my notes “nothing special.”

Then a talk about super-resolution for the Spitzer Space Telescope, essentially resampling and drizzling images taken at the same location to get a higher-resolution image. That was followed by lunch (yay) which was an entertaining Santa Fe chicken sandwich affair at a nearby supermarket.

Pan-STARRS came next, describing what the project is (wide-field survey with four little telescopes at each site) and how they were reinventing the analysis wheel YET AGAIN. How many people really need to re-roll their own data analysis software each time? It’s not like nobody’s ever written code to subtract two images before…

2MASS, the all-sky infrared survey, had a good talk (well, Roc Cutri gave the talk) about how they’re extending the mission and releasing new data products without needing to do any more observations. They’ve got a few patches of the sky that they’ve observed up to 3700 times, which means they can stack them all together to detect fainter objects than they detected earlier. This is kind of cool because some of the stuff I’m working on for WFCAM was going to use 2MASS to check results against. We’d been thinking that we wouldn’t need to use 2MASS because their data didn’t go faint enough, but now…

Anton Koekemoer was next with a talk about GOODS, the Great Observatories Origins Deep Survey. “Great Observatories” in this cases refers to three of the space-bound observatories — Hubble, Chandra, and Spitzer — along with some of the bigger and better ground-based observatories (including SCUBA on JCMT for the sub-millimetre regime). Just like anybody, they can present all kinds of really good scientific reasons why they need to do large-scale, hi-resolution, deep surveys at every wavelength possible. There were three other talks about space-based missions but there wasn’t much of interest in them. There do seem to be too many things called Gaia, though.

There were posters up all day but I got discouraged by the overzealous use of THE GRID in half of them. Mind you, I’m just as guilty as the next guy on this one, my poster has a data flow diagram that’s got a cloud with THE GRID label inside it. GRID GRID GRID can’t get enough.

There were a couple of BoFs (Birds of a Feather, essentially a “round table” discussion about a certain subject that isn’t much of a discussion at all because it usually gets taken over by the person with the microphone and that’s usually the last person who gave a 2 minute BoF presentation or the BoF chair), one about future astronomical data analysis environments where everybody seems really keen to re-invent the wheel, and one about Python where (I imagine, I didn’t actually go because what’s the point in a BoF about Python, where’s the Fortran BoF, where’s the C# BoF, where’s the Brainf*ck BoF) everybody seems really keen to re-invent the wheel in a completely different language. There were also some focus demonstrations where people show off a bit of software they’re writing and proud of (most of which is wheel re-invention — are you getting the point yet). These would be half-way interesting if the organisers had the foresight to imagine that perhaps more than 10% of people at a software conference would be interested in software and actually put the software demonstrations in a room that could hold more than 20 people. Brilliant, really.

But other than that it’s okay.