Denoising of crystal orientation maps

This paper compares several well known sliding-window methods for denoising crystal orientation data with variational methods adapted from mathematical image analysis. The variational methods turn out to be much more powerful in terms of preserving low-angle grain boundaries and filling holes of non-indexed orientations. The effect of denoising on the determination of the kernel average misorientation and the geometrically necessary dislocation density is also discussed. Synthetic as well as experimental data are considered for this comparison. The examples demonstrate that variational denoising techniques are capable of significantly improving the accuracy of properties derived from electron backscatter diffraction maps.