another restore function

[manono]

How would you guys go about this one?

Yadif(Mode=1,Order=0)#sample was BFF
MRestore(FRate=24)
#MRestore(Numr=2400,Denm=5000)#if using an older version, as I do

That works reasonably well.

[MOmonster]

Update:
Instead of mrestore, Cdeblend, FixBlendIVTC and DupHq there is only one function now: srestore.
The detection has been improved, the memory usage has been decreased and the usage should be a bit easier.
Just test this function.

Important:
- mrestore 2.31 has a big bug and has been removed from the R_pack
- AviSynth 2.5.8 change the calculation of changefps (round to floor). I will adapt srestore when the new AviSynth version gets final.

So, if you have problems with mrestore 2.31 or AviSynth 2.5.8 then test srestore with 2.5.7

@thetoof
Not sure what you really want. I could add a parameter for an alternative clip that is allways used if a blend is detected.
or do it this way:
line 123:
output.trim(0,1) ++ output.trim(2+current,0)
to something like this:
current==0 ? output : output.loop(3,0,0).trim(2+current,0).YourFilter(YourSettings)

Would be the same for srestore. If I wasn´t clear enough just ask.

Edit:
@Amrit
I used srestore with default settings on your source and don´t get any dups that aren`t in the source.

@vwpassion
Same here: Use your favourit bobber and srestore with default settings.

@all
If your problem is still not solved, please post it again. I don´t know anymore what problems are left.

[Adub]

Yes!! Thanks for your hard work MOmonster!!! Your script is really pretty amazing, and I use it too often. (grrr, bad sources = pure evil). This new single script will sure streamline alot of things, and I am sure it will make thetoof's life easier as well (as I use his script a bit as well).

[Amrit]

Thank you boss!!

[SweetDream]

Thank you MOmonster

[elguaxo]

Thanks!

[Piper]

@MOmonster: Thank you very much for this script! I've only just started using it but already getting very impressive results out of it. Well done! I think you've finally put ol' RePal out to pasture with this one.

[Deinorius]

Great, thanks. Will be interesting to test it soon.

[thetoof]

Wow, great job!

Actually, after doing some thinking, my request has changed a bit: could you add an argument that'd make srestore output a bool instead of a clip?
Something like bool "boolout" that would return true on blends and false on other frames (I'll use scriptclip). With this, I think I'll be able to get closer to what I'm trying to achieve.

[MOmonster]

@thetoof
I will see what I can do for you, but it has to wait over the weekend. Update with Avisynth 2.58 compatibility and so on will be next week.

[thetoof]

Thanks!
Another way of doing it that'd be convenient: a separate boolean function isblended() in srestore.avs

[MOmonster]

Update
I uploaded srestore v2.65. These are some of the changes:

- AviSynth 2.58 kompatibility
- light speed up
- reduced code
- new parameters: speed, thresh, pthresh, pclip
- better dup functionality

@thetoof
I don`t like the idea of calling a function that use the frame evaluation inside scriptclip. You can use this function as base. Rename it and add your own code:

function bcalc(clip source, float "speed", float "blocks", float "thresh", bool "chroma", int "cache", clip "dclip")
{
###### parameters & other necessary vars ######
	rad     = isfloat(speed) && speed>=1 ? sqrt(speed)*16 : 48
    cache   = default(cache,-1)
    bsize   = isfloat(blocks) && blocks>=1 ? sqrt(blocks)*4 : 16

global  thrx     = abs(default(thresh,22))*0.1
global  bfacx    = sqrt(bsize*0.25)*0.54


###### lut & more ######
detx	= default(dclip,source).killaudio().converttoyv12()
detx	= detx.pointresize(int(width(detx)/rad+1)*16, int(height(detx)/rad+1)*16)
detx	= default(chroma,false) ? stackvertical(stackhorizontal(detx.utoy(),detx.vtoy()),detx) : detx

    code	= "x 128 == y 128 == | x y + 256 == | 0 x 128 - abs y 128 - abs < x 128 - abs x y + 256 - abs < & x 128 - abs 1 - " +
            \ "y 128 - abs x y + 256 - abs < y 128 - abs 1 - x y + 256 - abs 1 - ? ? 2.4 ^ x 128 - y 128 - * 0 > -0.6 0.6 ? * ? 128 +"
	diff 	= mt_makediff(detx,detx.trim(1,0),y=3,u=1,v=1)
global  bclpx    = mt_lutxy(diff,diff.trim(1,0),expr=code,y=3,u=1,v=1).bilinearresize(round(width(detx)/bsize+4)*4,round(height(detx)/bsize+4)*4)


###### output preparation ######
global  outb    = blankclip(source)                                 #do your filtering here, it will be faster and more stable
global  outc    = source                                            #if you do it outside the conditional enviroment


###### initialise variables ######
global  lfrx     = -100


###### evaluation call & output calculation ######
scriptclip( source, """

 ### preparation ###
        cfrx = current_frame
        jmpx = lfrx+1==cfrx
    lfrx     = cfrx
        
        bx_v = 128-yplanemin(bclpx)
        bx_v = bx_v<0 ? 0 : bx_v * bfacx
        cx_v = yplanemax(bclpx)-128
        cx_v = cx_v<0 ? 0 : cx_v * bfacx
        
  ## blend value shifting ##
    bxp = jmpx ? bxc : bx_v
    bxc = jmpx ? bxn : bx_v
    bxn = bx_v
	
  ## clear value shifting ##
    cxp = jmpx ? cxc : cx_v
    cxc = jmpx ? cxn : cx_v
    cxn = cx_v

        isblend     = bxc+cxp>bxp+cxc+thrx && bxc+cxn>bxn+cxc+thrx                          #detects nearly every blend but also frames with very low motion
        isblendsave = isblend && bxc+cxp>bxp+(1+.2*thrx)*cxc && bxc+cxn>bxn+(1+.2*thrx)*cxc && bxc>(1+.2*thrx)*cxc     #saver blenddetection

 ### output clip ###
    isblend ? outb : outc                   #call your filtered sources here
    #outc.subtitle(string(isblend)).subtitle(string(isblendsave),y=16)      #to compare isblend and isblendsave

""")


###### final decimation & caching ######
cache<0 ? last : last.RequestLinear(8, cache, 5, false, false)

return last
}

[SweetDream]

Thanks for the new version

[Piper]

Yes, thank you for the new version. I'm getting some really nice results on some analogue captures with:

mvdegrain3i(2,0,1)
TDeint(order=1,field=1,mode=1,type=2,tryweave=true) 
crop(8,56,-8,-56)
NNEDI(field=1,dh=true,threads=2)
srestore()

Using TGMC would probably improve it further, but I'm not quite there with tweaking its denoising parameters... Still this is proving very nice. Speed is reasonable too. Thanks again.

[thetoof]

Thanks a bunch, this'll be very, very useful.

[EuropeanMan]

Am running my first srestore tonite...will report back as soon as it finishes; thanks MoM for all the hard work you've put in...very much appreciated.

Getting great speeds.

[MOmonster]

@all
Thanks for your kind words. Good to hear that the function is so useful.
I'm allready working on a new blenddetection to speed up srestore, but I don't when the next version will be released.

[ankurs]

thanks a lot !

[superuser]

thnxs a lot ... tried srestore works gr8.

though I was running into problems last nite using MT around srestore with 2.5.7 rev of Avisynth. Is MT recommend to use around SRestore for 2.5.7?

It was giving me error messages in the clip and can post SS for the same when I get back to my place.