When you first look at an image taken by the telescope, you are normally presented with a fairly disappointing looking black square with a few dim white dots. Such is the nature of astronomy cameras, imaging very faint objects using 16-bit image depth rather than the more usual 8-bit. This usually leaves the interesting image hidden deep within the data. After quickly ignoring the automatically generated JPEG on the "View" tab of the page, I'm sure we all go to the "Edit" tab, wait for it to load the real astronomy data file and play with the sliders on the left to find the target object lurking deep down in the levels. While easy to do by eye, this process is a little more tricky if you happen to be a web server. To have a computer automatically make a decent JPEG from an observation requires image processing algorithms to work on the file - and then hopefully create a good result.
We have had a few less than brilliant algorithms over the years which only really worked on brighter targets. The usefulness of the automatic JPEG creator was only ever supposed to be a quick attempt to make some sort of image - it was always expected that people would head straight to one of the editors. However, recently Ed proposed a new algorithm which produces better results.
So, after some really long-winded and dull sounding back-end work to route JPEG processing through our web processing software "Image Engine" and to have the job-view web page use our growing web API for creating images, etc etc, I present - better automatic JPEGs.
|NGC 2771 - old algorithm on the left, new on the right|
|An over-exposed Jupiter image suffers with the new algorithm|
|M20 appears from the new algorithm on the right|
One thing to note - if the web site finds that someone has already processed the image (with one of the older editors) and saved it to the gallery, it assumes that the human processed image will be better than the automatic and uses that instead.