Image Processing coding

Matlab BUG FOUND !!

  • if given an input of type single or double im2bw thresholds at x>T = 1 and x <T  =0, regardless of the range of the input variable x. T is however restricted to be between 0 and 1.
  • with uint8 and uint16, threshold T between 0 and 1 is simply 0 = 0, and 1 = 2^n where n is 8 or 16.
  • Of minor interest, B = im2bw(A,.2) is faster than B = A>.2*2^16, for thresholding uint16 for example.

New watershedding approach

Use max(I) -I as input instead of ~bwdist(~bw) .  Works pretty well.

Occassionally weak neighbor is fully in the ‘watershed’ of the stronger neighbor, which thus stays fused and the centroid gets mis-mapped.

 

 

New 3D dot finding:

Major changes to dotfinder to fix matlab bug.  Need to fix fxn_nuc_seg.m still.

3D dot positions (CheckDotUpDown.m).  Now constructs complete z-map for all positions inhabited by transcripts (might not scale well to large images, matrix is 2*Ndots x Zlayers, shouldn’t be too nonlinear in N-(possible)dots.

 

 

 

 

Can localize all dots in 3D.  Can we convert the pancake stacks into inferred real 3D gaussians?

 

 

This entry was posted in Microscopy. Bookmark the permalink.

Leave a Reply

Your email address will not be published. Required fields are marked *


*