Ventana BIF file format support

Benjamin Gilbert bgilbert at cs.cmu.edu
Sun Sep 22 20:20:22 EDT 2013


On 08/27/2013 06:01 AM, Agelos Pappas wrote:
> By trying to render tiles like in a regular TIFF files overlapping
> occurs. It is not clear to us yet how tile indexing works (left to right
> - top to bottom, spiral ?).

Tile indexing appears to be in row-major order with the origin at the 
bottom left and alternate rows reversed:

11 12 13 14 15
10  9  8  7  6
  1  2  3  4  5

With this interpretation, Tile2 is always directly UP/RIGHT of Tile1 and 
the recorded overlaps look reasonable.

> It also seems that stitching information is
> relative between tiles. This means that for different viewports where
> different tiles are visible, tile positioning will vary.

Also, we don't have valid alignment data for every pair of adjacent 
tiles.  We'll need to do global tile registration based on the available 
alignments and confidence values.  We also need this for Trestle (issue 
#21).

> Tile stitching information for the scaled down levels is not given.

In addition, the downsampled tiles have embedded overlaps, so we have to 
do MIRAX-style subtiling.

--Benjamin Gilbert



More information about the openslide-users mailing list