Further development of the method

The previous figures each come from the dissertation. Meanwhile, the fit process has been improved.

previous method advanced method
video

This results in a total of three advantages:

  • The fit process is accelerated.
  • The recognition rate of 96.8% of the pixels is increased to 99.96%, largely eliminating the need to “Repairing of small, non precisely fitted regions”.
  • The result will not darken the edges due to the motion correction.

    Originally, the fit program was written in Matlab on a workstation. Now I have rewritten it in C. It has the following runtimes:

    in Matlab on a workstation: 25–30 minutes
    in C on a Core2Duo (2.667 GHz): 3.1 seconds
    on a Core2Duo (2.667 GHz) with multithreading: 
    (ie distributing the work across both cores of the processor)    		1.9 seconds
    8 cores with 16 threads on a Intel i7-11700: 0.15 seconds
    GPU on the video card: not yet implemented

    In addition the efflux behaviour can be visualized. Here again, red means fast efflux and blue means slow efflux:

    visualization of efflux behaviour

    back