New halo removers discussion

[Didée]

The present haloing knocked me off the chair!

... But when hitting the ground, I had a raw idea worth to try:

rad = 3.0 # radius for halo removal
ss = 1.5 # radius for supersampling / ss=1.0 -> no supersampling
o = last
ox = o.width()
oy = o.height()

x = o.bicubicresize(m4(ox/rad),m4(oy/rad)).bicubicresize(ox,oy,1,0)
y = yv12lutxy(o,x,"x 8 + y < x 8 + x 24 - y > x 24 - y ? ? x y - abs * x 32 x y - abs - * + 32 /",U=2,V=2)
z1 = repair(o,y,1)
maxbig = y.expand().bicubicresize(m4(ox*ss),m4(oy*ss))
minbig = y.inpand().bicubicresize(m4(ox*ss),m4(oy*ss))
z2 = o.lanczosresize(m4(ox*ss),m4(oy*ss))
z2 = z2.logic(maxbig,"min",U=2,V=2).logic(minbig,"max",U=2,V=2).lanczosresize(ox,oy)

(ss == 1.0) ? z1 : z2

return last
#----------
function m4(float x) {return( x<16?16:int(round(x/4.0)*4)) }

No tweaks (could be done), no safety checks about ranges & so (probably needed), just your source image with raw halo press-down:

http://img157.imageshack.us/img157/9903/fuframe2904sourcedehalo0wt.png (http://imageshack.us)

[Chainmax]

I don't see that much haloing in the source frame (probably because it's so messy, I can provide cleaner frames with halos if you want) but I'm going to try it, thanks :).

[edit]Could you arrange it into a function? I don't know how to include it on the script.

[yaz]

vouw ... nother impressive script from didée :)
what i can't get :
- what is z1 ? ( i guess, it's the 'repaired' clip )
- why is 1.5 is hardcoded in the supersampling part ( i guess, that'd be ss instead )
... or am i on a wrong way ?

thx
y

[Didée]

- what is z1 ...
- why is ...
Dunno. Falling from the chair, I landed on my head ... straight thinking impossible.

Hey, it's just a sudden glance, quickly scribbled, and seems to do [at least] something right. I almost assure you there'll be some overflow-errors in the LUT ... can't evaluate that now.


For Chainmax:
function abcxyz(clip clp, int "rad", int "ss")
{
rad = default(rad, 3.0) # radius for halo removal
ss = default(ss, 1.5) # radius for supersampling / ss=1.0 -> no supersampling

ox = clp.width()
oy = clp.height()

x = clp.bicubicresize(m4(ox/rad),m4(oy/rad)).bicubicresize(ox,oy,1,0)
y = yv12lutxy(clp,x,"x 8 + y < x 8 + x 24 - y > x 24 - y ? ? x y - abs * x 32 x y - abs - * + 32 /",U=2,V=2)
z1 = repair(clp,y,1)
maxbig = y.expand().bicubicresize(m4(ox*ss),m4(oy*ss))
minbig = y.inpand().bicubicresize(m4(ox*ss),m4(oy*ss))
z2 = clp.lanczosresize(m4(ox*ss),m4(oy*ss))
z2 = z2.logic(maxbig,"min",U=2,V=2).logic(minbig,"max",U=2,V=2).lanczosresize(ox,oy)

return( (ss==1.0) ? z1 : z2 )
}

function m4(float x) {return( x<16?16:int(round(x/4.0)*4)) }

Obviously and as usual, RemoveGrain/Repair and MaskTools are needed.

[FredThompson]

Just for snicks and grins, look at the first post in this thread and see if enlarging, then noise filtering (maybe even a little heavier), then shrinking, then LimitedSharpen will help. http://forum.doom9.org/showthread.php?s=&threadid=68494 The sample images are gone but the results was wonderful. Maybe, just maybe, this would make the aberrations more aberrant...

FWIW, some folks combat that horrible blockiness in very similar areas by playback with ffdshow and adding a little noise. It can help break up the outlines just enough to fool our eyes during playback.

edit: Hah! I knew all this seemed too familiar: http://forum.doom9.org/showthread.php?s=&threadid=64432

[AVIL]

@Chainmax

AVIL: I used FFT3DFilter at sigma=6 and bt=3 (alone and with TemporalSoften) but it didn't do much to the rainbows.

Try with a script like :

avisource("your.avi")
fft3dfilter(sigma=3,sigma2=3,sigma3=12,sigma4=3,bt=3,plane=1)
fft3dfilter(sigma=3,sigma2=3,sigma3=12,sigma4=3,bt=3,plane=2)

In my case the rainbows are middle-sized so I use heavy denoising en sigma3. Real colour are more afected by sigma4, then I use a moderate value for it. Sigma and sigma2 copes whit subtle rainbows. You can raise this values but in the first example I think it isn't necessary.

Anyway, bifrost and smartssiq are very effective filters. In fact you can combine it:

From bifrost's manual.

Bifrost(scenelumathresh=1.5,altclip=SSIQ(11,100,true),interlaced=true)

But i found fft3dfilter more suited for me, in part for YUY2 support, in part for the easy in tunning the filter.

[Chainmax]

Didée, you are THE MAN!!! By replacing the FFT3DFilter line with a default call to abcxyz, the halos are completely gone and the result is almost identical to the last screen I posted!! :):). I'll post the three screens with this filterchain soon.

[FredThompson]

Didée does make some really good cleaning stuff. There are those, however, who think he's a shill for CPU manufacturers :P

Methinks I'll try this on some DV moire (camcorder shots of moving roller conveyor) and see how it does. Maybe it will also help with laserdisc moire.

[Chainmax]

Here are the three screens with the last filterchain:

http://img26.imageshack.us/img26/4499/finalscriptresult18rg.png (http://imageshack.us)

[link removed, screenshot appears now on page 3]

http://img470.imageshack.us/img470/7316/finalscriptresult34xt.png (http://imageshack.us)

The first one looks pretty much the same as the last attempt, but the second and third one look worse. They are, however, rare special cases that will probably be attenuated(sp?) by x264's deblocking (0,0 has proven too weak so I'll go for 2,2) anyway. I will encode a sample in a taxing sequence and upload it for you guys to examine. Again, thanks for all the input :).

[FredThompson]

You could play with MDeblock http://www.mdeblock.de.tf/

I like the results of LimitedSharpen but it's dog slow. The basic idea, modifying the extremes of the sharpening, could be put into dll sharpeners and they'd probalby run a lot faster. My Athlon XP 2200s run HC best profile at 5 fps with LimitedSharpen, ~30 fps without it. Something to keep in mind because it dramatically impacts encoding time.

[Chainmax]

Since these are 25min episodes, I can just compress to ffdshow's YV12 huffyuv and then feed that into MeGUI. I'll try MDeblock soon and report back, although it kinda bugs me that it never was developed over v0.3 and that it went pretty much unnoticed here.

[Didée]

@ Fred

Get this (http://forum.doom9.org/showthread.php?t=99679) and this (http://forum.doom9.org/showthread.php?p=731909#post731909), try Soothe(LimitedSharpen(ss_x=1.0,ss_y=1.0,Smode=4),last,24) [or 16], and compare again fps and result.

[FredThompson]

I'm now seeing 10 fps instead of 5 and the oversharp whities are reduced without losing the sharpness of the black lines. Thanks. DVD-RB Pro is now an overnight task, not a "sometime tomorrow" task. Thanks!!

What is the last parameter? I used your first suggestion, not the second. This is some sort of influence limiter?

The thread for Soothe mentioned the possibility of support files for LimitedSharpen. I wonder if anyone other than hanc messes with Fortran. Seems like a natural way to get some speed in the math routines.

Tried UnSharpMask instead of LimitedSharpen but the results just weren't that great. LimitedSharpen seems to do a very nice job of sharpening hard edges and leaving relatively smooth areas less modified. Quite helpful for animation, especially with the relatively low amount of oversharp artifacts.

[Chainmax]

I made a test encode of an expecially taxing sequence. You can download it from [link deleted].

[Anonymouses]

I'd just like to throw in that the dehaloing script you wrote Didée is pretty nice. Gets rid of the halos as good as BlindDeHalo but doesn't make the video blurry and doesn't thin out black lines.

Source:

http://img174.imageshack.us/img174/5405/source1qq.png

Using abcxyz():

http://img174.imageshack.us/img174/3412/abcxyz5jv.png

Using BlindDeHalo2(2.5,2.5,160):

http://img174.imageshack.us/img174/6596/blinddehalo22uh.png

And thanks for pointing out BiFrost Chainmax as it is a nice rainbow filter.

With BiFrost and abcxyz():

http://img174.imageshack.us/img174/1708/bifrost8pa.png

[Anonymouses]

I don't want to get too off topic here, but I can't get over how nice that dehalo script is. Here's another couple of samples comparing it against BlindDeHalo2 which I used to use. I changed the rad so that it takes float values.

Source:

http://img477.imageshack.us/img477/1320/sourcelupin2rc.png

default abcxyz():

http://img477.imageshack.us/img477/1926/abcxyzlupin7ry.png

abcxyz(rad=1.75), turned down the radius to preserve the small bright areas, and HQDering(255) to catch the bit of ringing that the lowered radius doesn't get:

http://img477.imageshack.us/img477/7838/abcxyzrad175lupin1ll.png

BlindDeHalo2(2.5,2.5,160):

http://img477.imageshack.us/img477/9868/blinddehalo2lupin2io.png

[Chainmax]

Very nice indeed. I'm going to test abcxyz (I renamed it to DDeHalo, i.é: Didée's DeHalo :)) on my test source that has lost of small details and see how it compares to HQDering, FixVHSOverSharp and BlindDeHalo3.

Anonymouses, I recently made a separate thread wher I ask for feedback on my Trigun encoding attempts. I'm mulling over two alternatives and would appreciate your feedback.

[Anonymouses]

Very nice indeed. I'm going to test abcxyz (I renamed it to DDeHalo, i.é: Didée's DeHalo :)) on my test source that has lost of small details and see how it compares to HQDering, FixVHSOverSharp and BlindDeHalo3.

Well at least in my tests I like to use both the dehaloer, with the radius turned down from default to stop it from messing with small bright areas, and using HQDering to pick up the haloing that is left behind due to the smaller radius. Though even at lower radiuses does get rid of more of the halo then just HQDering alone even when its set to 255.

[Anonymouses]

Chainmax just curious, how bad is the blocking on the R1 discs or is it pretty isolated? Looking at some of those shots with the pretty bad blocking and having seen how bad the rainbows are on the R1 discs, I'm glad that I bought the R3 boxset.

[Chainmax]

It's pretty nasty, but it is isolated too :p. It appears mostly on Vash's costume and a few rare times in sky scenes as well. I can upload a vob sample with the intro if you want to check it out.

[Anonymouses]

It's pretty nasty, but it is isolated too :p. It appears mostly on Vash's costume and a few rare times in sky scenes as well. I can upload a vob sample with the intro if you want to check it out.

Okay, I'm just curious to see how bad it really is. Though I think between the blocking and the rainbowing, that the rainbowing would probably more distracting from the screenshots I've seen.

[Chainmax]

Now that I look at the intro frame with the last filterchain, the white letters do look duller and yellowish. I'll try lowering DDeHalo's (the name I invented for abcxyz) radius to see if that makes them look better.

[Anonymouses]

Now that I look at the intro frame with the last filterchain, the white letters do look duller and yellowish. I'll try lowering DDeHalo's (the name I invented for abcxyz) radius to see if that makes them look better.

Yeah, I've noticed that lowering it to between 1.75 or 2.00 with HQDering picking up the slack looks better. Doesn't make the bright spots as greyish, and still does a good job of getting rid of the halos. Especially noticeable in the Castle of Cagliostro shots if you look at the chandelier.

[Chainmax]

Castle of Cogliostro? :confused:

Is that where the screens you posted in the 2nd page come from?

[Anonymouses]

Castle of Cogliostro? :confused:

Is that where the screens you posted in the 2nd page come from?

Yeah. Are you saying you've never heard of Castle of Cagliostro? http://www.iidb.org/vbb/images/smilies/eeka.gif

[Chainmax]

I am merely an ignorant newbie in the anime world. So far I've only seen (and sometimes not all of) DB, DBZ, Grave of the fireflies, Captain Tsubasa, Trigun, Saikano, Evangelion, Astroboy (80s methinks), Rurouni Kenshin, Spirited Away, Saint Seiya and a few others I can't remember right now. :o++

[Anonymouses]

I am merely an ignorant newbie in the anime world. So far I've only seen (and sometimes not all of) DB, DBZ, Grave of the fireflies, Captain Tsubasa, Trigun, Saikano, Evangelion, Astroboy (80s methinks), Rurouni Kenshin, Spirited Away, Saint Seiya and a few others I can't remember right now. :o++

Ahh you should definitely see it. It's one of Hayao Miyazaki's first films as director.

[Chainmax]

Let's discuss this further via PM :).

[Chainmax]

How could you set rad to 1.75? Its type is set to int, I had to switch it to float in order to be able to do that. By the way, 1.75 would be a much better default value, maybe toying with ss using HQDering as fallback could be avoided, can you please test that?

[Anonymouses]

How could you set rad to 1.75? Its type is set to int, I had to switch it to float in order to be able to do that.

Yeah, that's what I did. To quote myself:

I changed the rad so that it takes float values.

:)

By the way, 1.75 would be a much better default value, maybe toying with ss using HQDering as fallback could be avoided, can you please test that?

Sure, I can do that.

Edit: Messing around with the ss doesn't do anything. :(

[Anonymouses]

Actually been messing around with the radius again and I'm starting to like it around 2.25, a good balance IMO between halo removal and preserving the bright areas, and no need for HQDering. Heres some comparisons:

Source:

http://img476.imageshack.us/img476/5783/source3aa.png

Radius 3.00:

http://img476.imageshack.us/img476/4950/rad3001ou.png

Radius 1.75:

http://img476.imageshack.us/img476/7494/rad1754ob.png

Radius 2.25:

http://img476.imageshack.us/img476/9116/rad2256ds.png

Source:

http://img476.imageshack.us/img476/7011/sourcea4pm.png

Radius 3.00:

http://img476.imageshack.us/img476/696/rad300a4vv.png

Radius 1.75:

http://img476.imageshack.us/img476/613/rad175a2tu.png

Radius 2.25:

http://img476.imageshack.us/img476/7140/rad225a8ld.png

[Didée]

Guys. It's a trick function, aiming at removing brighties and preserving darkies. Reducing radius to preserve the brighties will also weaken the removal effekt - note that in your r=2.25 examples the halos are only damped, but not removed. While in the last shot, the letters still are noticeably darkened.

There were good reasons why initially I only posted a linear script, not a function ... and later named the function begged for as silly as "abcxyz": While it might have some potential, still it's only a trick function that *must* fail in places.


BTW, there was quite some mentioning of BlindDeHalo2 in this thread. Did anyone notice that BlindDeHalo has reached its 3rd incarnation since quite some time, and has gotten lots of tweaks to play with?

- "sharpness" parameter (and "tweak" counterpart)
- "lodamp" (and "hidamp" (!) ) parameter
- the "PP" modes, which act somewhat similar to HQdering ...
-- ... where "PP" < 0 runs *only* these routines, without the main dehaloing one

(a quickly fiddled before (http://img382.imageshack.us/img382/6382/vorher2uf.jpg) / after (http://img382.imageshack.us/img382/5219/nachher9ft.jpg) example, posted on the German forum some months ago.)

Know your tools. ;)

[Chainmax]

It might be a trick function, but it works wonderfully, and following rad=1.75 with HQDering is still a great option. Just because some people diss your awesome work that doesn't mean you must do it yourself. I really do believe that DDeHalo deserves its own thread and more testing by various persons.

About BDH3, I didn't mention it because I figured Anonymouses preferred 2 to 3. Besides, I just don't like the BDH line, I feel they oversmooth the whole frame and kill tiny details.

[Anonymouses]

Yeah but BlindDeHalo, even the 3rd incarnation, has a real problem with destroying black lines while this preserves them which is one things I like about it. Maybe something could be added to help preserve the black lines without needing to redarken them after using BDH? And true, while using a radius under 3.00 doesn't completely eliminate the halos it does dampen them enough that something like HQDering can be used to eliminate the last bit without hurting more of the bright areas and preserving the integrity of the black lines.

[yaz]

Guys. It's a trick function ... it might have some potential, still it's only a trick function that *must* fail in places ...yep yep ... make some tuning here and there and finally u get ... BDH ;) anyway. if u think of polishing abcxyz consider to move your resizing into the very last step a/o using spline16resize instead of lanczos. it'll fasten the encoding and decrease halo, definitely.

the bests
y

[Didée]

Oh, I think the resizing is already put as late as possible. And are you sure that spline16resize is faster and/or does create less ringing than lanczosresize?

No matter ... if I'd [manage to] put in the stuff exactly as you all would like, the speed would get very very slow anyway. We're probably talking about utilising median filtering with bigger radii, and such stuff. :)

[yaz]

@didée
i didn't mean u w/my comments but the abcxyz addicts :)

as regards spline16, yes, it produces much less ringing. speed? dunno. more or less the same. worth to give it a go.

what i meant by 'moving resizing in' is quite simple. chainmax's script contains a separate resizing step which is unnecessary if the downsizing step after the oversampling would target the final size. i would do it this way, but it's just a clue.

the bests
y

[Anonymouses]

yep yep ... make some tuning here and there and finally u get ... BDH ;)

Except BDH is much more destructive on details and the black lines in anime then the abcxyz function, so I'll just stick to using the trick function. :) Plus the speed is pretty good as well.

[Chainmax]

yep yep ... make some tuning here and there and finally u get ... BDH ;)

Your are wrong. Like anonymouses said, abcxyz has much less side-effects than the BDH line.
About Spline16Resize, is it as sharp as Lanczos4? Because if it isn't, then its speed and less likelyhood of creating halos is a moot point for me.


if I'd [manage to] put in the stuff exactly as you all would like, the speed would get very very slow anyway. We're probably talking about utilising median filtering with bigger radii, and such stuff

Like many others in this forum, speed is of no concern. The way I see it, it's better to spend twice the time if that will yield better results, after all you'll be seeing your encodes for far longer the time it took you to encode them. Again, I highly encourage you to make a thread for abcxyz and receive some extensive feedback from fellow Doom9ers. Animated content buffs will love it for sure.

[Revgen]

There is a lot of good information on dehaloing and ring removal in this thread, unfortunately this a rainbow removal thread.

Perhaps a mod could move all the dehaloing info to a new thread and a new topic.

[Chainmax]

Good idea, I'm going to email neuron2 and ask him to create a "New halo removers discussion" thread and move all posts from #21 onward to there.

[Guest]

Threads split as requested.

[Didée]

@ Revgen: Sorry for hijacking :o

@ neuron2: Thanks for splitting.


@ all: here's a new plaything :)

The following proof-of-concept function uses a method that I had in mind for a longer time already, but never actually tried it ... because during theorising I always thought "nah, this way it can't work. Well, I should have tried it earlier ... this is probably not the worst Halo remover ever seen :D

Currently, there are the following parameters:

- rx, ry [float, 1.0 ... 2.0 ... ~3.0]
As usual, the radii for halo removal.
Note: this function is rather sensitive to the radius settings. Set it as low as possible! If radius is set too high, it will start missing small spots.

- darkkstr, brightstr [float, 0.0 ... 1.0] [<0.0 and >1.0 possible]
The strength factors for processing dark and bright halos. Default 1.0 both for symmetrical processing. On Comic/Anime, darkstr=0.4~0.8 sometimes might be better ... sometimes. In General, the function seems to preserve dark lines rather good. ;)

- lowsens, highsens [int, 0 ... 50 ... 100]
Sensitivity settings, not that easy to describe them exactly ... in a sense, they define a window between how weak an achieved effect has to be to get fully accepted, and how strong an achieved effect has to be to get fully discarded.
Defaults are 50 and 50 ... try and see for yourself.

- ss [float, 1.0 ... 1.5 ...]
Supersampling factor, to avoid creation of aliasing.


source vs. DeHalo_alpha():

http://img366.imageshack.us/img366/6572/sourceagain6wp.th.png (http://img366.imageshack.us/my.php?image=sourceagain6wp.png) http://img366.imageshack.us/img366/323/dehaloalpha5tr.th.png (http://img366.imageshack.us/my.php?image=dehaloalpha5tr.png)


The function: (needs MaskTools and Repair from the RemoveGrain package)

(edit: forgot to include that small "m4" helper function. It's included now.)

(edit2: an "r" had disappeared, causing parameterless call to deliver a black clip. fixed :o )
function DeHalo_alpha(clip clp, float "rx", float "ry", float "darkstr", float "brightstr", float "lowsens", float "highsens", float "ss")
{
rx = default( rx, 2.0 )
ry = default( ry, 2.0 )
darkstr = default( darkstr, 1.0 )
brightstr = default( brightstr, 1.0 )
lowsens = default( lowsens, 50 )
highsens = default( highsens, 50 )
ss = default( ss, 1.5 )

LOS = string(lowsens)
HIS = string(highsens/100.0)
DRK = string(darkstr)
BRT = string(brightstr)
ox = clp.width()
oy = clp.height()
uv = 1
uv2 = (uv==3) ? 3 : 2

halos = clp.bicubicresize(m4(ox/rx),m4(oy/ry)).bicubicresize(ox,oy,1,0)
are = yv12lutxy(clp.expand(U=uv,V=uv),clp.inpand(U=uv,V=uv),"x y -","x y -","x y -",U=uv,V=uv)
ugly = yv12lutxy(halos.expand(U=uv,V=uv),halos.inpand(U=uv,V=uv),"x y -","x y -","x y -",U=uv,V=uv)
so = yv12lutxy( ugly, are, "y x - y 0.001 + / 255 * "+LOS+" - y 256 + 512 / "+HIS+" + *" )
lets = maskedmerge(halos,clp,so,U=uv,V=uv)
remove = (ss==1.0) ? clp.repair(lets,1,0)
\ : clp.lanczosresize(m4(ox*ss),m4(oy*ss))
\ .logic(lets.expand(U=uv,V=uv).bicubicresize(m4(ox*ss),m4(oy*ss)),"min",U=uv2,V=uv2)
\ .logic(lets.inpand(U=uv,V=uv).bicubicresize(m4(ox*ss),m4(oy*ss)),"max",U=uv2,V=uv2)
\ .lanczosresize(ox,oy)
them = yv12lutxy(clp,remove,"x y < x x y - "+DRK+" * - x x y - "+BRT+" * - ?",U=2,V=2)

return( them )
}

function m4(float x) {return(x<16?16:int(round(x/4.0)*4))}

[Chainmax]

At first glance DeHalo_alpha() seems just as good as abcxyz(rad=1.75) but with even a slight better treatment of small details: if you compare the abcxyz(rad1.75) picture Anonymouses posted with the DeHalo_alpha() one in the white beads chain that hold the chandelier, they seem a bit brighter on the DeHalo_alpha() one. Will test this beauty ASAP :).

[Dreassica]

Someone care to tell me which versiosn of masktools and removegrain i need, cuz i tried liek 3 different ones and all close my script upon loading.

[Didée]

MaskTools are these (http://manao4.free.fr/MaskTools-v1.5.8.zip) (unless I rebuild the script for MT v2.0aX)

RemoveGrain is not needed, but Repair.dll is. Latest official version should suffice for this one, although using the v1.0 pre-release (http://home.arcor.de/kassandro/RemoveGrain/RemoveGrain.rar) might be a good idea.


Oh, and:
At first glance DeHalo_alpha() seems just as good as abcxyz(rad=1.75) but with even a slight better treatment of small details: if you compare the abcxyz(rad1.75) picture Anonymouses posted with the DeHalo_alpha() one in the white beads chain that hold the chandelier, they seem a bit brighter on the DeHalo_alpha() one.
Yep, from the basic principle alone, this alpha should retain more detail than the abcxyz thingy.
However, when removing halos, there's always some apparent loss of overall contrast ... after all, the sharpening procedure that caused the haloing was not applied with the intention to bug the viewer, but to please him ;)

So, what about adding some minor areabased tralala afterwards?

http://img423.imageshack.us/img423/5281/dehaloalphasmallenhancement3mh.th.png (http://img423.imageshack.us/my.php?image=dehaloalphasmallenhancement3mh.png)
(quickly thrown together, so not tweaked at all.)

[Dreassica]

thnx, taht worked, but now using plain DeHalo_alpha() in script, i get all black frames.

[Didée]

thnx, taht worked, but now using plain DeHalo_alpha() in script, i get all black frames.
Nevermind - the halos are gone, aren't they? :D

Seriously: that's interesting! After finishing the function this morning, I also had the case that just a black clip was returned. Searched for an error in the script, but found none ... loaded the script again, and - ba-ding! - everything was fine.

Sorry, no clue ... try again.


edit: wait, something's fishy when defaults are produced. A moment ...

editedit: grrr, an "r" had disappeared somehow, and I didn't see it. ("brightstr" was "brightst"). Fixed. Sorry for being sloppy. :)

[Chainmax]

Yep, from the basic principle alone, this alpha should retain more detail than the abcxyz thingy.
However, when removing halos, there's always some apparent loss of overall contrast ... after all, the sharpening procedure that caused the haloing was not applied with the intention to bug the viewer, but to please him ;)

So, what about adding some minor areabased tralala afterwards?

http://img423.imageshack.us/img423/5281/dehaloalphasmallenhancement3mh.th.png (http://img423.imageshack.us/my.php?image=dehaloalphasmallenhancement3mh.png)
(quickly thrown together, so not tweaked at all.)

I don't know if I told you before, but You Are The Man!!.

I can't wait to try this one :).

[Anonymouses]

Didée, thanks for all the work you do and for refining this function more. Even better halo removing then with the original and preserves the small bright areas better.

Here's some more screenshots (especially make note of the eyebrows on the sculpture and the guys fingernails):

Source:

http://img313.imageshack.us/img313/9179/lupinsource4id.png

abcxyz(rad=1.75):

http://img313.imageshack.us/img313/4173/lupinabcxyz8bu.png

DeHalo_alpha():

http://img313.imageshack.us/img313/4651/lupindehaloalpha5tm.png

Wow, I think with this I can finally reauthor this disc without the halos and it preserves a great deal of the detail that BDH was removing.