Will Henney - DAWGI Meeting - 2019 June 24
Tetrabloks is an algorithm for making maps of inhomogeneous data. Different use cases for the algorithm include:
- Missing pixels: Tetrabloks will interpolate these regions away, even if they are large
- Noisy regions: Tetrabloks can smooth out high-noise regions while preserving the spatial resolution of low-noise regions
- Sparse and non-uniform spatial coverage: Tetrabloks can produce maps from an arbitrary set of points. It works well for combining multiple slit positions and orientations of longslit spectroscopy.
Multi-resolution mapping proceeds via two steps.
The first step is binning: each 2x2 block of pixels is averaged to give the next-coarser grid. Each pixel may have an associated weight, which is used in the averaging. A weight of zero indicates a missing or bad pixel (red in the figures), which does not contribute to the coarse grid. A tuneable parameter mingood
specifies the minimum number of good pixels that a 2x2 block must have in order to create a good pixel on the coarse grid. A value of mingood = 1
(the default) means that good regions "bleed" into bad regions, causing the bad regions to shrink. The binning is repeated to produce a sequence of coarser and coarser grids: 1x1, 2x2, 4x4, 8x8, etc.
The second step is stack the sequence of grid. In the simplest version (as illustrated), each pixel comes from the finest grid where it has a non-zero weight. Alternatively, other criteria can be used, such as a minimum signal/noise ratio.
Before.
After.
Individual binning levels.
Fixed signal-to-noise of the density.
Inspiration for the name.