suomela / mf2d Goto Github PK

Is the current code build for the iOS platform ?

Dear all,

i am writing this more of a warning to fellow googlers who find themselfs here; there is really not too many float32 / uint16 libraries that could do median sorting with large window (like 9x9 or 21x21 - as required on certain standard).

This library seemed very fine, and during course tests it seemd to provide correct results. When studying with detail due some other issue, i did notice bit weird output. As attachement is the processed image from the examples compared to ImageJ "median filter" output of radius=2.0.

The m2fd output has to my untrained eye kind of "liney artefacts" that would seem wrong to my intuition. I have not done pen&paper verification to this.

On left is the m2fd output, on right is the ImageJ median filtered and on the right is the un-altered data. The gray-levels have been adjusted to same on each image.

When doing image subtraction (m2fd output image - imagej median filter r=2 image) and compare the results are following:

Left: The m2fd output and its histogram (to give some idea of the leves) and on right the difference image (and its histogram for the levels).

The command to obtain the result image was

./bin/mf2d 2 ./example/test-1.fits /tmp/out.fits /tmp/out-diff.fits

The platform is linux x86_64 (kubuntu 20.04)