XuvTools is developed in a cooperative effort from:
  • Chair of Pattern Recognition and Image Processing [www]
  • Friedrich Miescher Institute for Biomedical Research [www]
  • Center for Biological Systems Analysis [www]
 

Improving Data Structures

The Stitcher does not currently employ the concept of data bundeling. Every date is in an individual variable (possibly n-dimensional, though). It would be appropriate to bundle several dates that are commonly used together in algorithms.

The data type 'tileXXXX' is used tightly with:

  • tile coordinates (currently separate structure)
  • tileXXXX_enabled
  • tileXXXX_channel
  • stackXXXX with blitz++ image data

The data type 'pairwise_displacements' is often used to compute 'tile coordinates'. The data type 'pairwise_displacements' is used tightly with

  • 'pairwise_correlation_coeff'

Pairs of Tiles

Pairs of tiles (PoT) are a cornerstone of the concept of stitching. Here is a collection of information that a PoT can be assigned:

  • Is a displacement known (multiple combinations of these are possible):
    • Is a displacement known from stage coordinates? This will not have a correlation coefficient assigned.
    • Is a displacement known from manual pre-alignment? This will not have a correlation coefficient assigned.
    • Is a displacement known from phase-only correlation? This will have a correlation coefficient assigned that has no absolute meaning, and is useful only in relation to other correlation coefficients of the same pair.
    • Is a displacement known from normalized cross-correlation? This will have a quantitative correlation coefficient assigned.
  • Are both tiles currently in RAM available?

Multiple combinations of known displacements are possible. For some displacement types, a list can emerge: i.e. several positions can have been cross-correlated with different correlation coefficients.

devel/improving_data_structures.txt ยท Last modified: 2010/04/10 10:36 by mario
Contact: admin(a)xuvtools.org