Dear list members,
I made some ExtendedPluginFilters from the fast block filters available
in the mpicbg library. The filters use the integral image trick to
calculate the sum in a rectangular block in constant time regardless of
the block's size. That way we have implemented
Mean,
(Sample)Variance,
Standard Deviation,
normalized cross correlation,
and mean-scaling
We have used the STD and mean filter for statistical outlier removal. A
pixel is considered an outlier if it is larger or smaller than
mean+-a*STD
in a block centered around the pixel. a and the block-size are
parameters of the filter. Outliers are replaced by smearing in
surrounding non-outlier pixels.
Impressive with the filters is the speed when used with large
block-sizes, please use the preview option.
Although integral image filters show their real strength when used for
calculating the result for many block-sizes on demand (which can be done
using the mpicbg API), the filter plugins here may already be useful for
daily use when time plays a role.
Find the filters in Fiji under
Plugins->Integral Image Filters
or in a package for standalone ImageJ including classes, sources and a
few interactive demonstrations:
http://fly.mpi-cbg.de/saalfeld/download/integral_image.jarBest regards,
Stephan
Locked
