BlindDeHalo() - removal of Edge Enhancement artefacts

[dvwannab]

Originally posted by State of Mind
Can someone give me a direct download link for this filter, please? I wish to test it on some DV footage.

Hi State of Mind!! Glad you took my advice and come over here. These guys are fantastic arent they? This is place to play with, test and learn avisynth. I am gonna be testing this new filter also.
Good luck. To use BlindHalo2, you have to copy the code (link above) and paste into wordpad and save as blindhalo2.avs. Then use import command. Ex.

Loadplugin("C:\avisynth\filters\xxxx.dll")
Import("C:\avisynth\filters\blindhalo2.avs")

Also need MaskTools 1.5.5.

[State of Mind]

To use BlindHalo2, you have to copy the code (link above) and paste into wordpad and save as blindhalo2.avs. Then use import command. Ex.

What link?

[Socio]

Originally posted by SirCanealot
Has anyone got any suggested settings or advice for using this on anime?
I can get it to remove SOME of the halos, but it does as much damage to random high-contrast parts of the picture as it does EE removal.

Thanks all.

Try the first BlindDeHalo Didee posted, top page #1 instead of BlindDeHalo2.

Originally posted by State of Mind
What link?

Soulhunter's post at the bottom of the first page.

[Soulhunter]

Originally posted by ObiKenobi

Soulhunter's post at the bottom of the first page.

Bottom of 2nd page...

EDIT: At least with the default "Posts per Thread" settings !!!


Bye

Originally posted by Soulhunter
Bottom of 2nd page...


Bye

Well for me it's the first page.

[Soulhunter]

Originally posted by ObiKenobi

Well for me it's the first page.

Ok, Ive edited my last post...


Bye

[Blue_MiSfit]

Awesome script didee! Works extremely well on the film "Human Stain" that displays horrendous edge enhancement haloing.

BlindDeHalo(2,2,100)
Preserves sharpness much better than
UnFilter(-20,-20)

Well done, though it took me awhile to get the proper versions of the BlindDeHalo / MaskTools in order Speed isn't bad at all, 6-10fps on my A64 @ 2.34GHz with a 960x480 (lanczos4resized) encode in the firstpass for xvid 1.1.

Very nice

~misfit

[BlackMetal]

Hi, I'm trying to use BlindDeHalo2, but I get the following error:
Script error: there is no function named "overlay"

I've tried with MaskTools 1.5.1, 1.5.5, 1.5.6 and -p4-5.
Does anyone know what's wrong?

[Manao]

It's not a filter of the masktools, but of avisynth. What is your avisynth version ?

[BlackMetal]

Originally posted by Manao
It's not a filter of the masktools, but of avisynth. What is your avisynth version ?

Yes, you were right. I had avisynth v2.5.2.0.
Now I tried with v2.5.5.0 and it worked.
Thanx!

[Socio]

Originally posted by Blue_MiSfit
Awesome script didee! Works extremely well on the film "Human Stain" that displays horrendous edge enhancement haloing.

BlindDeHalo(2,2,100)
Preserves sharpness much better than
UnFilter(-20,-20)

~misfit

Hey that gives me an idea;

Didee, what would happpen if you replace Unfilter in your IIP script with BlindDeHalo2?

[Didée]

Nothing spectacular would happen.

Remember that iiP offers two different methods for deringing: its "old" routine (called with positive "dering" values) - that's the one which is using "unfilter" -, and the "new method" (called with negative "dering" values), which most times works better in iiP's context.

BlindDH2() is effective against Halos, but doesn't do all too much about mosquito noise. The "new" deringing method of iiP is better suited as "general purpose" deringing filter, as it's designed to counteract both types of artefacting.

So ... if at all, then it's iiP's method that should be merged into BlindDeHalo(), not the other way round


Drifting from topic, that's how LimitedSharpenEX() will be alike: It will clobber you over the head with ... possibilities

On the de-ringing/de-haloing side, you'll have a variety of choices - 2 or 3 main types of operation, each coming along with 2 or 3 sub-types of slightly different strengths or characteristics.
Then there will be an asymmetric sharpener available, especially for VHS processing, but perhaps also for aspect conversions 1:1 <--> anamorhic, back'n forth.
Not to mention that the LimitedSharpen principle will be extended by a detail crystality enhancer, an area adaptive contrast booster, and optional color enhancement. Also, some easy implementations of gauss/sharp-mixing will be in.
But there was something more, whatwasit, whatwasit ... ah yes, that little "noise-bypass-sharpening" functionality ...

As can be seen, a bunch of topics that arose in this forum in the last few weeks, have already been on my mind when I started the project last autumn ... "Didée, master of sluggish development"

(If only I'd find more time to work on it, again. The finishing state drags out like chewing gum, but I won't release some half-finished alpha-thingy of such complexity.)

[Macanudo]

Originally posted by Didée

Drifting from topic, that's how LimitedSharpenEX() will be alike: It will clobber you over the head with ... possibilities

On the de-ringing/de-haloing side, you'll have a variety of choices - 2 or 3 main types of operation, each coming along with 2 or 3 sub-types of slightly different strengths or characteristics.
Then there will be an asymmetric sharpener available, especially for VHS processing, but perhaps also for aspect conversions 1:1 <--> anamorhic, back'n forth.
Not to mention that the LimitedSharpen principle will be extended by a detail crystality enhancer, an area adaptive contrast booster, and optional color enhancement. Also, some easy implementations of gauss/sharp-mixing will be in.
But there was something more, whatwasit, whatwasit ... ah yes, that little "noise-bypass-sharpening" functionality ...
[/B]

Stop...You are making me drool...(insert picture of Homer Simpson staring at a box full of donuts)...LOL.

[Socio]

Originally posted by Macanudo
Stop...You are making me drool...(insert picture of Homer Simpson staring at a box full of donuts)...LOL.

I say lock Didee in a room til it's finished!

[dvwannab]

Originally posted by Didée
Nothing spectacular would happen.

Remember that iiP offers two different methods for deringing: its "old" routine (called with positive "dering" values) - that's the one which is using "unfilter" -, and the "new method" (called with negative "dering" values), which most times works better in iiP's context.

BlindDH2() is effective against Halos, but doesn't do all too much about mosquito noise. The "new" deringing method of iiP is better suited as "general purpose" deringing filter, as it's designed to counteract both types of artefacting.

So ... if at all, then it's iiP's method that should be merged into BlindDeHalo(), not the other way round


Drifting from topic, that's how LimitedSharpenEX() will be alike: It will clobber you over the head with ... possibilities

On the de-ringing/de-haloing side, you'll have a variety of choices - 2 or 3 main types of operation, each coming along with 2 or 3 sub-types of slightly different strengths or characteristics.
Then there will be an asymmetric sharpener available, especially for VHS processing, but perhaps also for aspect conversions 1:1 <--> anamorhic, back'n forth.
Not to mention that the LimitedSharpen principle will be extended by a detail crystality enhancer, an area adaptive contrast booster, and optional color enhancement. Also, some easy implementations of gauss/sharp-mixing will be in.
But there was something more, whatwasit, whatwasit ... ah yes, that little "noise-bypass-sharpening" functionality ...

As can be seen, a bunch of topics that arose in this forum in the last few weeks, have already been on my mind when I started the project last autumn ... "Didée, master of sluggish development"

(If only I'd find more time to work on it, again. The finishing state drags out like chewing gum, but I won't release some half-finished alpha-thingy of such complexity.)

WOOOHOOOO baby!!!! Bring it on..........bring it on

Just wanna add that I love BlindDeHalo2 and can't wait for you to bring out LimitedSharpenEX.

These two shots are from Episode 10 of Trigun from the R2 DVDs

Source:



Filtered with BlindDeHalo2(2.5,2.5,160) & LimitedSharpen(ss_x=2.0,ss_y=2.0,Smode=3,strength=100):



And the second example:

Source:



Filtered with BlindDeHalo2(2.5,2.5,160) & LimitedSharpen(ss_x=2.0,ss_y=2.0,Smode=3,strength=100):

[dvwannab]

ATM,

Nothing meant by my comments, but first I didnt see anything wrong with your source and the applied filters gave a washed-out look. Loss in contrast and saturation. Dont see how LS & Blind helped at all. Actually, they took away. May want to tweak your settings or just not do anything at all

Originally posted by dvwannab
ATM,

Nothing meant by my comments, but first I didnt see anything wrong with your source

You mean other then the strong EE haloes?

Originally posted by dvwannab
and the applied filters gave a washed-out look. Loss in contrast and saturation. Dont see how LS & Blind helped at all. Actually, they took away. May want to tweak your settings or just not do anything at all

They took away the EE and nothing else. Neither BlindDeHalo or LS have any effect on contrast or saturation, and those were the only filters applied. So I have no clue what you are referring to. The EE gives a perceived increase in contrast, but that is completely artificial (same way that it gives a perceived higher level of detail when in fact its just the opposite).

Here are the two histograms based on the 2nd screenshots.

*Removed pics see below*

There is a bit of difference in the luma channel but only at the outlying parts (due to the EE haloing being removed). In the color channels there is no real change at all. So I'm not sure how you can say there is washed out colors when the color channels haven't been changed to any noticeable extent.