Dogway's Filters Packs

[Dogway]

poisondeathray: I will have a look. Actually I had plans to replace GrainFactory3mod completely as it shouldn't be that slow. I want to make use of the Addgrain corr args to avoid resizing clip 3 times.

Julek: Do you have an input and output sample? I got some code but I'm not sure what I should target. Also there are adaptive mean/median/gaussian, which one do you need?

[gispos]

Then please also check ex_contrast, parameter "sat" does not work for me.
Thank you.

[Dogway]

poisondeathray: I just checked and it's fine here, using your colorbars call and the GrainFactory3mod call from OP. The call is actually "GrainFactory3mod" not "GrainFactory3" (fixed in OP). Might you be missing a dependency?

gispos: I tested and yes, ex_contrast() defaults to UV=1 for YUV sources. I fixed it by enabling UV=3 if sat is defined, will upload for next update.

UV    = Default(UV, Defined(sat) || rgb ? 3 : 1)

In any case the saturation of ex_contrast() isn't meant as a full fledged saturation function, that's the task of ex_vibrance() but I admit it's a bit abandoned lately. Among other things I also need to make it color space aware. ex_contrast() saturation is a simple lerp, more to keep along luma contrast.

[poisondeathray]

Originally Posted by Dogway

poisondeathray: I just checked and it's fine here, using your colorbars call and the GrainFactory3mod call from OP. The call is actually "GrainFactory3mod" not "GrainFactory3" (fixed in OP). Might you be missing a dependency?

Yes, GrainFactory3mod . I mentioned GrainFactory3 because that separate older (non ex) function works

If I copy/paste the function into the script and rename it something like GrainFactory3mod2 , it still returns original clip, so it's not autoloading some other function with the same name

Yes all dependencies are met on the list, with same or newer versions

[gispos]

Originally Posted by Dogway

gispos: I tested and yes, ex_contrast() defaults to UV=1 for YUV sources. I fixed it by enabling UV=3 if sat is defined, will upload for next update.

With UV=3 it works, I hadn't even looked at UV.

[Dogway]

poisondeathray: Maybe you have a function (name) duplication? Another thing to care about avs+ version. Also are you using MIX mod? Both should work though but I always use MIX mods myself mostly. In any case download again from the repo (just updated) and check for dependencies.

ExTools updated to v4.6.
-ex_blur() whole refactor. Now it uses true gaussian coefficients and float radius. For matching removegrain(12), blur(1) use ex_boxblur(mode="weighted")
-ex_smartblur() bugfix that prevented the function from working
-ex_binarize() dynamic mode
-ex_median() optimize "medianST" mode

EX/MIX mods
-Rebased to latest ExTools
-Reverted removegrain(19) to removegrain(20)

I also tried to port removegrainHD smartmedian modes but failed to it, spent whole day yesterday. Might add my attempts if anyone is interested to give it a stab.

[poisondeathray]

Originally Posted by Dogway

poisondeathray: Maybe you have a function (name) duplication? Another thing to care about avs+ version. Also are you using MIX mod? Both should work though but I always use MIX mods myself mostly. In any case download again from the repo (just updated) and check for dependencies.

AVS+ 3.7.1 r3431

Yes I just re-downloaded and check prerequisites. Not sure what's going on

Neither GrainFactory3mod EX or MIX mods work for me. I renamed the functions to call them separately. I even removed autoloading GrainFactory3mod.avsi versions and manually copy/pasted function into script . I double check that function is missing - ie. error message when GrainFactory3Mod is called

eg. rename function

function GrainFactory3mod_ex(clip...

function GrainFactory3mod_mx(clip...

They return the original clip for me

The original GrainFactory3mod from 2020.04.06 (I think real_finder) works ok

[Dogway]

I don't know, there's nothing special on my mod. In these situations I would work my way up within the function to check up where it starts to fail. Output process clip in different stages to see what's failing. Usually it tends to be a dumb oversight. Use simple script, empty the plugin folder, etc.

If you can please try replacing the last line:

use_ALMask ? mt_merge(result...  }

With just

result }

[poisondeathray]

Originally Posted by Dogway

I don't know, there's nothing special on my mod. In these situations I would work my way up within the function to check up where it starts to fail. Output process clip in different stages to see what's failing. Usually it tends to be a dumb oversight. Use simple script, empty the plugin folder, etc.

If you can please try replacing the last line:

use_ALMask ? mt_merge(result...  }

With just

result }

Same thing;

It's ok - Don't worry about it, the original mod still works ok for me

[Julek]

Originally Posted by Dogway

Julek: Do you have an input and output sample? I got some code but I'm not sure what I should target. Also there are adaptive mean/median/gaussian, which one do you need?

Here is an example of mean and gaussian with blockSize=15 and C=3, both applied to luma after a prefilter.
I usually use the mean option more.
Link

[Dogway]

ExTools updated to v4.7.
-Guidelines for common operations with ex_xxpand(): Opening, closing, smooth, outline, inline, gradient, top hat, black hat.
-ex_hitormiss(). New function for advanced binary mask operations. Modes: Thin, thicken, prune, pruneS, boundary, convexhull, convexhull45, corner, cornerS, skeleton, SKIZ, end and junction. I may add more in the future.

EX/MIX mods
-Minor cosmetic stealth updates

[Dogway]

@Julek: I tried to port adaptive threshold, there are many flavours though but couldn't make anyone work, unless my integral image is wrong I think the problem is in the second part, the thresholding.

It involves two passes, since the integral overflows range_max I'm testing in 32-bits first.

# mean of the local sum
str = "x[-2,-2]  A^ x[-1,-2]  B^ x[0,-2]  C^ x[1,-2]  D^ x[2,-2]  E^ 
       x[-2,-1]  F^ x[-1,-1]  G^ x[0,-1]  H^ x[1,-1]  I^ x[2,-1]  J^ 
       x[-2,0]   K^ x[-1,0]   L^ x[0,0]   M^ x[1,0]   N^ x[2,0]   O^ 
       x[-2,1]   P^ x[-1,1]   Q^ x[0,1]   R^ x[1,1]   S^ x[2,1]   T^ 
       x[-2,2]   U^ x[-1,2]   V^ x[0,2]   X^ x[1,2]   Y^ x[2,2]   Z^ 

                 A B +      WA^    WA C +         WB^   WB D +           WC^   WC E +           WD^
    A F +  WE^   F G WA + + WF^    F G H WB + + + WG^   I WG WC + + WB - WH^   J WH WD + + WC - WI^
    K WE + WJ^   K L WF + + WK^    K L M WG + + + WL^   N WL WH + + WG - WM^   O WM WI + + WH - WN^
    P WJ + WO^   P Q WK + + WP^    P Q R WL + + + WQ^   S WQ WM + + WL - WR^   T WR WN + + WM - WS^
    U WO + WT^   U V WP + + WU^    U V X WQ + + + WV^   Y WV WR + + WQ - WX^   Z WX WS + + WR - WY^

    A WA WB WC WD WE WF WG WH WI WJ WK WL WM WN WO WP WQ WR WS WT WU WV WX WY + + + + + + + + + + + + + + + + + + + + + + + + 25 /"

For the second part I'm not sure whether I should pick the box "vertices", or their range, and if the bottom-right corner should be [x,y] or [x+2,y+2]. In any case I tested all possible combinations (5x5 only, not 15x15 yet) but didn't yield any meaningful results.

Some attempts for thresholding:

str = "x y[2,-2] y[-2,2] + y[-2,-2] - y[2,2] - 25 / 1 0.14 - * <= 0 range_max ? "
str = "x y[2,2] y[-2,-2] + y[-2,2] - y[2,-2] - 25 / 0.8 * <= 0 range_max ? "  # this seems legit
str = "x y[0,0] y[-2,-2] + y[-2,0] - y[0,-2] - 9 / 1 0.04 - * <= 0 range_max ? "
str = "x[0,0] y[0,0] y[-2,2] + y[-2,0] - y[0,2] - 9 / 1 0.91 - * <= range_max 0 ? "
str = "y[0,0] y[-2,2] + y[-2,0] - y[0,2] - 9 / 0.003 <= range_max 0 ? "  # this seems legit
str = "x y[0,0] y[-1,0] y[-1,-1] y[0,-1] y[-2,-2] + + + + y[-2,0] - y[-2,-1] - y[0,-2] - y[-1,-2] - 9 / 1 0.958 - * <= range_max 0 ? "
str = "x[0,0] y[0,0] y[-1,0] y[-1,-1] y[0,-1] y[-2,-2] y[-2,0] y[-2,-1] y[0,-2] y[-1,-2] y[-2,-2] + + + + + + + + + y[-2,0] - y[-2,-1] - y[0,-2] - y[-1,-2] - 9 / 1 0.9939 - * <= range_max 0 ? "
str = "x         y y 25 / A@ - B@ 1 B - / 1 - 0.445 * 1 + A *      <= 0 range_max ? " # y should be local sum (without mean)
str = "x         x y 25 / A@ - dup * sqrt 128 / 1 - 0.34 * 1 + A *      <= 0 range_max ? " # Sauvola (y is local mean -no integral-)
str = "x         x y 25 / A@ - dup * sqrt 0.5 / 1 - 0.25 *   -10 A * exp 2 * 1 +   +   A *      <= 0 range_max ? " # Phansalkar (y is local mean -no integral-)

I might need to research a bit more.



By the way, for Otsu's I found a way to create histogram bins, well a poor man's version of it, simply threshold image for each bin, subtract the previous thres and average each bin get the amount of values.

ScriptClip("""
thr1=ex_binarize(25.5,true)
bin1=thr1.AverageLuma()
thr2=ex_logic(thr1, ex_binarize(51,true), "andn")
bin2=thr2.AverageLuma()
... """)

[Dogway]

ExTools updated to v4.8.
-ex_binarize() added "otsu" mode.
-ex_median() fixed "adaptive" (for good). Optimized "median5", "medianST", "medianT" and "medianT5"
-ex_repair() optimized "median5"
-ex_bs() upgraded to convert from any bitdepth to any bitdepth, including self bitdepth with different bit scale.
-Updated benchmarks

EX/MIX mods
-Rebase to latest ExTools

SMDegrain updated to v3.2.4d.
-Revert to removegrain(20)
-Fix ex_BM3D() issue when input is 32-bit
-Revert workaround for empty 'threads' issue
-Swapped mt_merge with ex_merge, thus totally removing masktools2 dependency

GradePack updated to v2.6.
-Update ex_bs()
-Optimize ex_blend() "interpolation" mode
-ex_contrast() enable chroma if sat is defined.

[Dogway]

@Tormento: Couldn't refrain myself so I cobbled this up in a moment. Defaults are not tuned, check against STPresso() to see how it compares, this runs about 100fps, STPresso 3.3fps.

# Spatio-Temporal Thresholded Weighted Median

function STTWM(clip a, int "sthres", int "tthres", int "athres", int "slimit", int "tlimit", int "UV", bool "fulls") {

    rgb = isRGB(a)
    isy = isy(a)
    bi  = BitsPerComponent(a)

    ths  = Default(sthres, 24)          # 0 - 100%
    tht  = Default(tthres, 49)          # 0 - 100%
    tha  = Default(athres,  1)          # 0 - 100%
    sl   = Default(slimit,  5)          # 0 - 255
    tl   = Default(tlimit,  5)          # 0 - 255
    UV   = Default(UV,    rgb ? 3 : 1)
    fs   = Default(fulls, rgb)

    ts   = ths/100.
    tt   = tht/100.
    ta   = (100-tha)/100.
    rt   = 1 - (ts + tt) * ta

    Assert(ths+tht<101, "Spatial and Temporal thresholds cannot sum more than 100%.")
    Assert(IsVersionOrGreater(3,7,1), "Update AviSynth+ version to 3.7.1+")

    str =  Format(                                         \
           "x[-1,1]  x[1,1]       dup1 dup1 min K^ max P^
            x[-1,-1] x[1,-1]      dup1 dup1 min L^ max O^
            x[0,1]   dup                        D^     N^
            x[-1,0]  dup                        C^     M^
            x[0,-1]  dup                        I^     J^
            x[1,0]   dup                        G^     H^
            x[0,0]   dup dup dup                A^     F^
                                                B^     E^
            N P                   dup1 dup1 min N^ max P^
            F O                   dup1 dup1 min F^ max O^
            J M                   dup1 dup1 min J^ max M^
            I L                   dup1 dup1 min I^ max L^
            B K                   dup1 dup1 min B^ max K^
            E H                   dup1 dup1 min E^ max H^
            D G                   dup1 dup1 min D^ max G^
            A C                   dup1 dup1 min A^ max C^
            H P                   dup1 dup1 min H^ max P^
            M O                   dup1 dup1 min M^ max O^
            E N                   dup1 dup1 min E^ max N^
            C L                   dup1 dup1 min C^ max L^
            G K                   dup1 dup1 min G^ max K^
            F J                   dup1 dup1 min F^ max J^
            A I                   dup1 dup1 min A^ max I^
            B D                   dup1 dup1 min B^ max D^
            O P                             min O^
            L N                   dup1 dup1 min L^ max N^
            H M                   dup1 dup1 min H^ max M^
            J K                   dup1 dup1 min J^ max K^
            D I                   dup1 dup1 min D^ max I^
            F G                   dup1 dup1 min F^ max G^
            C E                   dup1 dup1 min C^ max E^
            A B                                    max B^
            M O                             min M^
            K N                             min K^
            H L                   dup1 dup1 min H^ max L^
            G J                   dup1 dup1 min G^ max J^
            E I                   dup1 dup1 min E^ max I^
            C F                                    max F^
            B D                                    max D^
            K M                             min K^
            E L                   dup1 dup1 min E^ max L^
            H J                   dup1 dup1 min H^ max J^
            G I                   dup1 dup1 min G^ max I^
            D F                                    max F^
            K L                             min K^
            I J                             min I^
            G H                                    max H^
            E F                                    max F^
            I K                             min I^
            F H                                    max H^
            H I                   dup1 dup1 max swap2 min

            x swap2 clip S@

                     dup dup dup
            a dup
            z dup
            y b                   dup1 dup1 min swap2 max
            swap4  swap1  swap8   dup1 dup1 min swap2 max
            swap4  swap1  swap6   dup1 dup1 min swap2 max
            swap9  swap1  swap3   dup1 dup1 min swap2 max
            swap7  swap1  swap5   dup1 dup1 min swap2 max
            swap2  swap1  swap4   dup1 dup1 min swap2 max
            swap3  swap1  swap6   dup1 dup1 min swap2 max
            swap2  swap1  swap7   dup1 dup1 min swap2 max
            swap4  swap1  swap8   dup1 dup1 min swap2 max
            swap9  swap1  swap3             min
            swap8  swap1  swap5   dup1 dup1 min swap2 max
            swap2  swap1  swap4                       max
            swap2  swap1  swap7   dup1 dup1 min swap2 max
            swap7  swap1  swap4   dup1 dup1 min swap2 max
            swap2  swap1  swap5   dup1 dup1 min swap2 max
            swap6  swap1  swap3   dup1 dup1 min swap2 max
            swap2  swap1  swap7             min
            swap3  swap1  swap4   dup1 dup1 min swap2 max
            swap4  swap1  swap5   dup1 dup1 min swap2 max
            swap2  swap1  swap6                       max
            swap3  swap1  swap2             min
            swap3                 dup1 dup1 min swap2 max
            swap2  swap1  swap4                       max
            swap2                           min
            swap2                                     max

            x swap2 clip T@
              x - abs {tl} > x T ? {tt} *
            S x - abs {sl} > x S ? {ts} * + {ta} * x {rt} * + ")

        b  = a.selectevery(1,-1)
        f  = a.selectevery(1,+1)
        b2 = a.selectevery(1,-2)
        f2 = a.selectevery(1,+2)

        isy     ? Expr(a, b2, b, f, f2, str,                                  optSingleMode = false                     ) : \
        UV == 1 ? Expr(a, b2, b, f, f2, str, "",                              optSingleMode = false                     ) : \
                  Expr(a, b2, b, f, f2, str, ex_UVexpr(str, UV, bi, rgb, fs), optSingleMode = false, scale_inputs="none")

}

[tormento]

Originally Posted by Dogway

Couldn't refrain myself so I cobbled this up in a moment.

I suffer OCD too

Originally Posted by Dogway

Defaults are not tuned, check against STPresso() to see how it compares, this runs about 100fps, STPresso 3.3fps.

I just want to replicate

Spresso(10,30).STpresso(10,30,4,22,8,49,1,mc=true)

just a bit faster

[Dogway]

I'm trying to port mt_hysteresis() (slow and seems a little buggy on each F5) to ExTools. I'm still not sure this can be ported even by reading the source code.

I managed to get a very similar output with this, which basically is just a 8-connected hit structuring element. The other blocks in red still don't add or help at all.

function ex_hysteresis(clip a, clip b, int "UV", bool "fulls") {

    rgb = isRGB(a)
    isy = isy(a)
    bi  = BitsPerComponent(a)

    UV = Default(UV,    rgb ? 3 : 1)
    fs = Default(fulls, rgb)

    str = "x[-1,-1] A^ x[0,-1] B^ x[1,-1] C^
           x[-1,0]  D^ x[0,0]  E^ x[1,0]  F^
           x[-1,1]  G^ x[0,1]  H^ x[1,1]  I^
           y[-1,-1] J^ y[0,-1] K^ y[1,-1] L^
           y[-1,0]  M^ y[0,0]  N^ y[1,0]  O^
           y[-1,1]  P^ y[0,1]  Q^ y[1,1]  R^

           J A min
           K B min max
           L C min max
           M D min max
           N E min max
           O F min max
           P G min max
           Q H min max
           R I min max X^
             
           J 
           K max 
           L max 
           M max 
           O max 
           P max 
           Q max 
           R max  Z^

           J N min
           K N min max
           L N min max
           M N min max
           O N min max
           P N min max
           Q N min max
           R N min max Y^

           Y 0 != range_max 0 ? "

    isy     ? Expr(a, b, str,                                     optSingleMode=false) : \
    UV == 1 ? Expr(a, b, str, "",                                 optSingleMode=false) : \
              Expr(a, b, str, ex_UVexpr(str, UV, bi, rgb, fs), optSingleMode=false) }

Reading the VS docs says:

"If we treat the planes of the clips as representing 8-neighbourhood undirected 2D grid graphs, for each of the connected components in clipb, the whole component is copied to the output plane if and only if one of its pixels is also marked in the corresponding plane from clipa."

The X^ block tries to compare every corresponding pixel from clipb to clipa, so if it's a hit, then the Y^ 'connected' block can be evaluated and passed to output clip. I also tested by simply comparing E to Y.

[Pauly Dunne]

Hi Dogway,

I just wanted to say that I have been using SMDegrain & ExTools (early builds) successfully in RipBot 264, but with the lasted builds I get an error when running AVSMeter saying there is an issue with Line 446 in SMDegrain.

Are there any other dependancies that need to be used with the later builds ???

I can work around by using older builds of SMDegrain.

Regards

[Dogway]

Maybe if you are doing contrasharpening with values you need my mod of LSFmod, or if you are using BM3D prefiltering you might need Utils-r41.
I noticed that in latest LSFmod I changed 'defaults' arg to 'preset', try changing that in line 337 until I update SMDegrain.

[Pauly Dunne]

Originally Posted by Dogway

Maybe if you are doing contrasharpening with values you need my mod of LSFmod, or if you are using BM3D prefiltering you might need Utils-r41.
I noticed that in latest LSFmod I changed 'defaults' arg to 'preset', try changing that in line 337 until I update SMDegrain.

Thanks for the info, and yes, one of the scripts that throws up the line 446 error IS a contrasharp=30 entry...

Also, I checked the version of LSFmod I was using, and it's was a little old @ v2.187, the latest that I could find was v2.191...doesn't that sound right ??

I just had a look at Line 337 in v2.191, it's blank

Don't know anything about BM3D....is there a forum or something around ??

I'll do some updates, and let you know

OK, back already, and with the very quick response from "videoh" & "kedautinh12", I have some stuff to study thanks guys

So I now have LSFMod v2.9mx, so I will see what that does.

[videoh]

Originally Posted by Pauly Dunne

Don't know anything about BM3D....is there a forum or something around ??

https://forum.doom9.org/showthread.p...highlight=BM3D

Read the whole thread. test3 supports AVS+.