halfQuadraticFilter not smoothing?

[vedadb]

• Having applied the filter on the same EBSD maps several times, sometimes it filters and sometimes not: o Can not reproduce conditions for either case o The success of the smoothing is here quantified by the mean KAM over the EBSD map

• In the following I show: o Comparison between KAM maps of the largest grain of two EBSD samples (no. 1 and 11, see below) with no filtering, hQF and spline filter on. o This effect was first noticed on full maps but is here illustrated on single grains o A summary over 11 EBSD maps comparing no, hQF, and spline filtering; all other filters (mean, median, Kuwahara, etc.) have been checked that they also smooth the data but are not included here.

• Attached with the document is a .zip file containing the @EBSD map of the single largest grain, without filtering, from two samples (No.1 and No.11) and a script file to reproduce the figures: o no1_largest_grain.mat & no11_largest_grain o first cell in issues_script.m hQF_issues.zip

1a. Summary

Figure 1. Mean KAM plotted for 11 samples without filtering, with hQF (F.alpha=1) or with spline filtering. Samples 1 to 5 and 10 show no effect of smoothing by hQF, while all KAM-values are lowered upon spline filtering (smoothed orientation maps)

1b. Specific KAM maps

Figure 2. Comparison between filtering methods on the KAM maps for Sample 1 and 11. Sample 1 hQF is obviously not being smoothed properly.  

[kilir]

Hi, what alpha parameter do you use? The default doesn't do a lot. in the current version, values are NOT normalized to be in the range [0 ,1] and maybe something around 100-300 should give reasonable results. Cheers, Rüdiger

[kilir]

Hi, this was an issue for some input data, fixed here fb40dea, so e.g. using ebsd.quality = ebsd.quality ./ max(ebsd.quality(:)) should correct the input and reasonably small alpha values should work again. Cheers, Rüdiger

[vedadb]

Hi, thank you, that fixed it. For one of the grains I needed an alpha of 10000 to reach similar smoothing result as alpha = 1 for the other, now they are normalized.

Vedad