Information on Service 'Dexter for your data'

[Use this service from your browser]

Further access options are discussed below

An interface to ADS' Dexter data extraction applet for usage with uploaded data

For a list of all services and tables belonging to this service's resource, see Information on resource 'dexter/ui'

Service Documentation

Dexter supports you when extracting data from scans of printed material or other bitmap images. It is not designed as a source extractor from sky images. There are much better programs doing that.

Dexter is implemented as a Java applet. Its use is documented elsewhere.

Here are some usage hints on the standalone version:

Giving names

When going to the service's entry page, you will be asked for a name. You can freely choose one. If you do not delete your results after work, you and others can come back later to retrieve your results or do a second extraction [1] by giving this name.

Uploading the source

You can then upload data (if the name was already taken, you will see the page previews of the document already present under the name). You can upload PDF documents or images in GIF, PNG, and JPEG formats. Dexter was designed to operate on scanned images, so its input data always has to be a relatively high-resolution bilevel (i.e., black/white) image. For PDFs, dexter will compute these images from your input.

For bitmap images like jpeg files, this is a process with many degrees of freedom. We let you influence some of these in the "Assume DPI" and "Binarization method" fields (which are ignored for PDF input).

Under "Assume DPI", you can give the "scan resolution" of your image. The effect is that it will be scaled up by 600/<assumed DPI>. Of course, we cannot add detail that is not in the image, but scaling up helps with the representation of greyscale in the bilevel image, so in particular for colour input you will want to put something like 200 or so here. Do not overdo it, though, since many Java VMs built into browsers will crash your browser if asked to load giant images.

The binarization method can be simple thresholding (everything that has more than half the maximal intensity becomes white, everything else black) or one of several error diffusion methods. Try Floyd-Steinberg first if your image looks really bad in Dexter. You can experiment by re-uploading with various settings.

Starting the extraction

Once the data is in, you will see previews for the page(s) on the top of the resulting web page (and the option to re-upload below [2]) as well as a selection of data sets that may have been extracted before.

Click on the page you want extracted. If your input was a PDF, you will arrive at Dexter's area selection. For images, we assume you already more or less selected the area of interest and send you to Dexter's extraction window right away, with the entire image. The area selection widget is still present, albeit very small.

From that point on, let the Dexter documentation help you. We may in particular mention the recognizers that frequently help you assign axes, or find points and trace lines for you.

[1]Dexter currently cannot load existing results. We could retrofit this functionality if there is some demand, so just send us a mail if you need this.
[2]Also note the link "Delete this data set from server" at the very bottom of the page; use this if you do not want other people from being able to guess the name you gave and see your data.


You can access this service using:

This service is published as follows:

local means it is listed on our front page, ivo_managed means it has a record in the VO registry.

VOResource XML (that's something exclusively for VO nerds)