Conditional Temporal Median Filter
http://kevin.atkinson.dhs.org/temporal_median/

Version 0.93 for AviSynth 2.5
September 27, 2003

I've been waiting for this one since 20.06.2003 when i asked kevin if he could make an avs or vdub versions of his CTM filter.....he said he couldn't do avs or vdub filters, but in the meantime he did "avisynth_c.dll" which acts as a go-between and connects kevin's programming to avs.........

after a problems with initial releases,let's see what can this baby do to remove spots that look like this;

well,see if this one looks any better;

it DOES look better,doesn't it?

this test uses the same methodology as explained in my declick filter.htm test [both declick and CTM attempt to do the same thing->to remove the "clicks" {small dots and streaks} from the damaged video]

actually,to make it more interesting i have even taken the screenshots of same frames that were done with declick (to make it direct comparison), naming methodology is the same too,only here "wdr-ctm-xxxx.jpg" means the ConditionalDenoise was applied before deinterlacing (ie. on interlaced stream) and "wdr-ctm-xxxx-progressive.jpg" means CD was applied after the deinterlacing (ie. on progressive stream).......

i suggest you collect all the images from this test and mentioned declick filter.htm test, put them all in one folder and cycle to see what are the differences between these two filters.....

ok,here are the screenshots....

wdr-ctm-0006-progressive.jpg
wdr-ctm-0029.jpg
wdr-ctm-0029-progressive.jpg
wdr-ctm-0035.jpg
wdr-ctm-0043.jpg
wdr-ctm-0044.jpg
wdr-ctm-0044-progressive.jpg
wdr-ctm-0056.jpg
wdr-ctm-0069.jpg
wdr-ctm-0075.jpg
wdr-ctm-0075-progressive.jpg

every image is cca.85kB

so here's how i tested;
768x576(mjpeg)->CD->kerneldeinterlaced->bicubicresize(512,384,0,0.75)->"wdr-ctm-xxxx.jpg"
and
768x576(mjpeg)->kerneldeinterlaced->CD->bicubicresize(512,384,0,0.75)->"wdr-ctm-xxxx-progressive.jpg"

script:
#ConditionalDenoise(pwidth=32,pheight=3,interlaced=true,p1=64,p2=32,mthres=48)
or
#ConditionalDenoise(pwidth=32,pheight=3,interlaced=false,p1=64,p2=32,mthres=48)
in latter case...
(these setings were suggested by kevin himself after he inspected this video....as at that time i couldn't play with the filter myself because it still had some problems...now i see his setting works very nice so for such source i won't even try to tweak it....)

btw.now i can do that vhs-rip i was talking about.....

copyleft 27.09.2003. by i4004:everyone is invited to make some more tests of this filter

contact: kosta_@net.hr