Fixing and adding things onto a new DotCrawlPlus version

[RGMOfficial]

Hi, guys.

So, since the original DotCrawlPlus is a little less than 10 years old, it's expected to become somewhat obsolete.

I'm thinking of forking this filter, after all, this filter has last updated in 2014, and many things changed, specially on my perception of analog video artifacts in general.

So, yeah, i'm starting making an unofficial version 3.
After all, some parts of the filter are not as realistic as actual analog video footage (like this comment).

On the moment, my missions on this project are:
- Improvements in performance
- Moving rainbows
- Better rainbowing
and some missions i'm starting to think otherwise...

I'm still new to some stuff, so any help are appreciated.

I'm starting with the moving rainbows thing.
Just need to know if i can animate mt_lutspa or something like that?

Also, credits to @raffriff42, the nice guy who created the original filter.

[SaurusX]

I fully endorse your plan. Degradation filters for AVISynth are in short supply and they would definitely be of use on occasion. Training neural networks for instance.

When it comes to fake composite rainbows, there's been some discussion on one of the Discord channels I'm on by using the following:

AddGrain(0,0.6,0.4,50000,seed=1) #credit to @Zarxrax

Overlaying it with a diagonal edge mask gives a decent approximation.

[real.finder]

maybe someone can use https://forum.doom9.org/showpost.php...9&postcount=19 to make a plugin for avs

[RGMOfficial]

Originally Posted by SaurusX

I fully endorse your plan. Degradation filters for AVISynth are in short supply and they would definitely be of use on occasion. Training neural networks for instance.

When it comes to fake composite rainbows, there's been some discussion on one of the Discord channels I'm on by using the following:

AddGrain(0,0.6,0.4,50000,seed=1) #credit to @Zarxrax

Overlaying it with a diagonal edge mask gives a decent approximation.

So, i tried.

Changed the makedotcrawl function, but maintained the mask.

##################################
### generate a rainbow pattern
##
## @ sat    - chroma intensity (default 1.0)
## @ xluma  - ratio of clip cross-luma to fixed pattern (0..1; default 0.75)
## 
function dpc_makerainbow(clip C, float "sat", float "xluma")
{
    sat  = Float(Default(sat, 1))
    xluma = Float(Min(Max(0, Default(xluma, 0.75)), 1))
	
    C
    AddGrain(0,0.6,0.4,50000,seed=1)
}

Test images on attachments (using rainbow=1.0)

[RGMOfficial]

Did some changes on the code, and actually...

the rainbowing looks great and realistic.

function dpc_makerainbow(clip C, float "sat")
{
    sat  = Float(Default(sat, 1))
    
    color_sat = int(50000*sat)
	
    C
    AddGrainC(var=0,uvar=color_sat,hcorr=.9,vcorr=.4,seed=1)
}

Also, i added a new preset, that i've been using with the original plugin. It's the preset used on the screenshot below.

## @ preset - choices:
##            "mild"     - minimal effect
##            "medium"   - softer chroma; a little added noise
##            "heavy"    - very soft & noisy
##            "strong"   - like medium but more so
##            "bigdot1"  - larger dots #1
##            "bigdot2"  - larger dots #2
##            "bigdot3"  - larger dots #3
##            "betacam"  - Analog formats - Betacam
##            empty string ("") defaults to "medium";
##            unrecognized values raise an error.

[SaurusX]

Another thing to consider regarding rainbows is that they generally live only on DIAGONAL edges. Would it be preferable to alter the mask to only affect diagonal edges? It would add another layer of realism to the effect.

[RGMOfficial]

Originally Posted by SaurusX

Another thing to consider regarding rainbows is that they generally live only on DIAGONAL edges. Would it be preferable to alter the mask to only affect diagonal edges? It would add another layer of realism to the effect.

Can you provide an example script?

Here's the mask script:

function dpc_uumidpass(
\   clip C, int "tlo", int "thi", bool "setupmode")
{
    tlo    = Default(tlo, 128)
    thi    = Default(thi, 148)
    setupmode = Default(setupmode, false)

    CG = C.Grayscale()
    Last = CG
    Blur(1, 1)
    FFT3DFilter(bt=1, bw=64, bh=64, ow=32, oh=32,
    \  sigma=0, sigma2=0, sigma3=500, sigma4=100000)
    Subtract(UUGaussianBlurY(rad=1))
    mt_expand()
    Grayscale()
    return (setupmode) 
    \ ? dpc_histostretch(50.0, 0.01, 1.8, true)
    \ : Levels(tlo, 1.8, thi, 0, 255, coring=false)
}

function UUGaussianBlurY(clip C, float "var", float "rad")
{
    var = Max(0.0, Float(Default(var, 1.0)))
    rad = Max(1.0, Float(Default(rad, Pow(var, 0.5))))
    var = Pow(Min(Max(0.0, rad), 60.0), 2.0)
    
    w0 = C.Width
    h0 = C.Height
    w1 = Round(w0/rad)
    h1 = Round(h0/rad) /* formerly "w0/rad" ?? */

    B = C.BilinearResize(
    \         Min(Max(12, w1 + (w1 % 2)), w0), 
    \         Min(Max(12, h1 + (h1 % 2)), h0))

    B = B.Blur(1.0).Blur(1.0)
    
    return (var<0.01) ? C 
    \    : (var<1.5)  ? C.Blur(0.85 * var)
    \    : (B.Width>8 && B.Height>8) ? B.GaussResize(w0, h0)
    \    : B.BilinearResize(w0, h0)
}

Also, sorry for the late reply.

[ognirats]

Holy hell that looks nice

[ognirats]

Now how do I implement it

[ognirats]

Nvm, found this thread https://forum.doom9.org/showthread.php?t=184188
it works wonderfully