ucl-bug/ust-sart

Code for reconstructing 2D sound speed maps from time-of-flight data using straight-ray-SART

ust-sart's Issues

Timing

Implement timing tracking so that bottlenecks can be addressed

Vectorised pixel weight calculation

The reconstruction time is dominated by the calculation of pixel weights for each ray section, which currently uses a for loop. Need to vectorise this to speed it up:

Time for breaking rays into sections: 0.90166 s
Time for calculating pixel weights:   66.2126 s
Total time:                           70.6406 s

remove k-wave toolbox dependency

manually implement getWin and waterSoundSpeed

Spiking at reconstruction circle

Spiking errors are observed at the reconstruction boundary, in the region known to be water.

Possible causes:

• rays that intersect the reconstruction circle almost tangentially, meaning residual errors are distributed over only 1 or 2 grid points.
• upsampling the grid causing the reconstruction circle to be defined differently (a point previously excluded from the recontruction circle becomes included, and has a large sound speed contrast relative to surrounding points)

Possible solutions:

• Implement maximum value changes
• Change the size of the reconstruction circle after upsampling
• Smooth the correction matrix before applying it.