Preparing the pixels

At Zoo headquarters we like to be efficient. That means avoiding redoing work that has already been done by someone else. Particularly if those others have already spent a long time thinking how to do it best. Getting the images for the original Galaxy Zoo (way back in 2007!) was particularly easy. The fabulous Sloan Digital Sky Survey (SDSS) had already done all the work of taking the images, calibrating them, stitching them together, combining images at different wavelengths to make colour images, and optimising their appearance. All we had to do was ask their servers for an image, giving it the required location and size, and voilà, a image ready for adding straight into the Galaxy Zoo collection!

Life was rather more difficult when we added the special ‘Stripe 82’ images from the SDSS. For these, Galaxy Zoo team member Edd needed to do the stitching, combining, optimising, cutting-out and resizing. The details of how he did that are all here. We wanted to be able to compare the Stripe 82 images to the normal SDSS images, so we tried to keep things like the brightness scaling and appearance of colours as similar to the original as possible. Even so, it took us a couple of attempts to come up with a solution we were satisfied with.

With the Hubble data, as with Stripe 82, creating the images for the Zoo isn’t completely straightforward, but again most of the hard work had already been done for us. For the launch of Galaxy Zoo: Hubble, data was taken from several surveys:

• GOODS: The Great Observatories Origins Deep Survey
• GEMS: Galaxy Evolution from Morphology and SEDs
• AEGIS: All-wavelength Extended Groth strip International Survey

We’ve also recently added in COSMOS: Cosmic Evolution Survey images – more about the nitty gritty details of those images in a future post.

The data calibration business was already taken care of by the science teams for each survey. The next steps, finding the galaxies, cutting out images at each available wavelength and combining them into colour images, was handled by Roger Griffith, who already had a system set up to do exactly that. Roger used a nifty piece of software called GALAPAGOS to manage the business of finding, cutting out and measuring the galaxies. The difference that Galaxy Zoo added to Roger’s system was that, like with Stripe 82, we wanted the properties of the colour images to match those from SDSS as closely as feasible, to enable us to compare the results from each of the Galaxy Zoo datasets as fairly as possible.

One particular issue with making colour HST images is that many surveys only produce data at two different wavelengths. Normally, colour images are made by choosing a different wavelength image for each of the three primary colours: red, green and blue. For the HST images we instead use one image for red, another for blue, and then just take the average of the two for green. The primary colours used in your computer display don’t usually match the colour filters that were used in the telescope at all, so the colours you see are only an indication of the true colour. Nevertheless the colours contain a lot of information: galaxies containing only old stars will look red, while those which are actively forming new stars will often be blue. Getting the images looking right, with fairly similar appearance to the SDSS images, required a cycle of testing and exchanging images back and forth, but we came to an agreement fairly quickly.

The HST images in Galaxy Zoo might not look as impressive as some of the press images you’ve seen from Hubble over the past twenty years. That’s because press images are usually picked specifically for their attractive appearance. The images chosen are often of nearby nebulae and galaxies for which HST allows us to see huge amounts of detail. The objects in Galaxy Zoo: Hubble are much more typical of the huge number of galaxies in HST surveys. Although HST can see much more detail than ground-based surveys, its mirror and field-of-view are smaller than most ground-based telescopes, so it can only cover a much smaller area of the sky in a reasonable amount of time. Surveys with the HST therefore focus on faint, distant galaxies, so we end up with images having similar quality to those from SDSS, which is remarkable given how much further away the HST galaxies are compared with those from SDSS.

The similarity between the images of galaxies in the early universe from HST and those relatively nearby from SDSS is actually a big advantage. It means that we can fairly compare the morphologies of galaxies at these two eras in the Universe’s history. That’s what professional astronomers will be doing with your Galaxy Zoo: Hubble classifications over the coming year.