Technical details - Acohir

Why not use quicktime-VR?

We wanted higher resolution than QTVR can cope with - typically up to 3kx2k per image view. This means using some multiresolution image format or compression and not sending all the data in one go. We also wanted more control over colour, although in the end we chose sRGB, which could also be used in QTVR, we could do more colour calibration in the Java client.

Our client-server approach is clearly slower than QTVR (once that has loaded everything - which can be slow!) in use but is more extensible. It also allows the viewer to show wha the object looks like by only fetching icon-level data.

How does it work?

Captured images are colour calibrated to CIE Lab using a colour chart.
They are stored as sRGB in a TIFF file which is tiled, pyramidal, JPEG.
These provide 64x64 image tiles at multiple resolutions in JPEG format from the image server CGI.

Each of the 20-40 views is a separate TIFF file. Typically 2-3MB each, so 40 views is ~ 100MB.

The Java browser uses the Internet Imaging Protocol to fetch each tile. So as the scrollbar is moved, or the mouse drags the large view, new tiles are fetched from the server. These requests look something like this:

http://server/iipserver.fcg?FIF=monalisa.tif&OBJ=IIP,1.0&JTLS=5,4,3

IIP standardises fetching parts of images from a web server, so it allows multi-vendor clients and servers.
 
There are two Java applets, one for the icon and the other for the main view. The Zoom/etc controls underneath are JavaScript as are the view links in the text on the right.

 
 

Back to ACOHIR page