New Script: Drawrect 0.6

[jmac698]

@Fluff
You have to realize that Avisynth script is, by far, the easier way for me to create specific videos as I require. I am not a Windows programmer and do not know enough Video For Windows to create a new video, and anyhow I still think it would be tremondously complicated to try to learn that now. Otherwise, I need to research the AVI file format, come up with a header, research the raw format, and do the byte packing or complicated calculations to draw directly into Y1 U Y2 V bytes in 32 bit words, which are then padded. Not simple at all.

The only other way I know of is the math expressions feature in ImageMagick.

Please, tell me even one method which is simpler than what I'm doing, which won't take an hour of reading API's, or file formats.

[jmac698]

@Fluff
http://msdn.microsoft.com/en-us/library/ms779636.aspx

So I will actually try what you suggest.
I should write :
'RIFF' (filesize) ('AVI')
ckID ckSize
...
I also have to create an AVI index... there really seems to be a lot of research and calculations involved. Do you see my point?

[TheFluff]

Originally Posted by jmac698

@Fluff
You have to realize that Avisynth script is, by far, the easier way for me to create specific videos as I require. I am not a Windows programmer and do not know enough Video For Windows to create a new video, and anyhow I still think it would be tremondously complicated to try to learn that now. Otherwise, I need to research the AVI file format, come up with a header, research the raw format, and do the byte packing or complicated calculations to draw directly into Y1 U Y2 V bytes in 32 bit words, which are then padded. Not simple at all.

The only other way I know of is the math expressions feature in ImageMagick.

Please, tell me even one method which is simpler than what I'm doing, which won't take an hour of reading API's, or file formats.

When I said an uncompressed video I meant raw containerless RGB or YUV data, not an AVI.

Anyway, while there is certainly no shortage of sane(r) ways to do what you want, the obvious answer here is to write an Avisynth plugin. Blending a single-color rectangle on top of an existing image is literally about fifteen lines of rendering code in addition to the maybe 50 lines of Avisynth boilerplate stuff that you can just copypaste from somewhere. There is not exactly a shortage of good sample code, since most plugins are open source.

The entire GetFrame() function in the original assrender (my first and so far only Avisynth plugin), which essentially blends RGB32 bitmaps with a given opacity and a given pixel offset on top of an RGB32 image, is about 60 lines of code total, which includes about 10 lines of comments. The inner rendering loop is 17 lines. Go take a look at lachs0r's C implementation (which takes care of some really complicated shit like rendering RGB32 on top of YV12 with chroma interpolation and all, without converting the source image like overlay() does) and you're basically almost done already. It's a sunday afternoon project, really.

[jmac698]

@Fluff
Ok, I tried another possibility. You can draw with imagemagick. Great. But I still can't specify YUV colors:
http://www.imagemagick.org/script/co...ions.php?#fill

You are right, there's lots of easy ways to draw RGB shapes, but nearly impossible to draw YUV colors.

So I looked into converted raw files.

convert -size 720x480 movie.yuv

Should work. I still haven't found the specification for YUV raw file format, however.

Originally Posted by TheFluff

Also, you should obviously write an Avisynth plugin. ... There is not exactly a shortage of good sample code, since most plugins are open source.

Again, I've been trying to do just that. The result is my article, http://avisynthnew.wikinet.org/wiki/...to_your_Plugin

I've had to give up, however, as I was not able to modify the sample code to work. The best I'd be able to do is not change a thing except for the pixel calculation part, thus resulting in a plugin with useless parameters. And I really doubt there's a plugin with just enough int's in it's parameters to cover my needs.

It's not as simple as copy & paste, though part of it is. Parsing and checking the parameters takes some though of course.

Perhaps you could write a sample pure C plugin that does nothing but average two clips, then I could base my article off that. Please release as GPL.

[Gavino]

Originally Posted by jmac698

I've had to give up, however, as I was not able to modify the sample code to work.

What was the problem?
You had this more or less written almost a year ago and I pointed out the errors then.
What more do you need to know?

[jmac698]

Thanks for asking, I'll start a new thread.

[TheFluff]

Originally Posted by jmac698

Should work. I still haven't found the specification for YUV raw file format, however.

There is no such "file format", you just dump raw YUV data to disk with any plane order you like. Of course you have to know that order as well as the resolution and subsampling mode in order to make any sense of it afterwards.

Originally Posted by jmac698

Perhaps you could write a sample pure C plugin that does nothing but average two clips, then I could base my article off that. Please release as GPL.

I have no interest in your project (beyond questioning its sanity). Furthermore, I don't have any experience at all with the avisynth_c interface, so you probably know more about it than me.

[Dogway]

@jmac698, I don't know why you quote me so late in time, as I'm already using the nice lutspa function you gave me.

function BoxMask(clip clip, int "x1", int "x2", int "y1", int "y2", bool "show"){

w=clip.width()
h=clip.height()

x1=string(default(x1,0))
x2=string(default(x2,w))
y1=string(default(y1,0))
y2=string(default(y2,h))
show=default(show, false)

mt_lutspa(clip, mode="absolute", yexpr="x "+x1+" >= x "+x2+" <= & y "+y1+" >= y "+y2+" <= & & 255 0 ?", chroma= "128").trim(0,-1).FreezeFrame(0, FrameCount(clip)-1, 0)
show?mt_lutxy(clip,last,"x 0.75 * y 0.25 * + 2 /"):last}

For fancy shapes masks I normally paint it myself and load it as imagesource.

[jmac698]

@Fluff
There are pre-determined YUV formats which ImageMagick will accept. I think the convert option only accepts 1:1:1 sampling. I'm still uncertain if it expects planar or interleaved, and/or if I can specify which to use. I certainly do not see any command line options with the words "planar" in them.

Anyhow, thanks for your suggestions. I would desperately love to be able to write plugins, and have started another thread on that topic, perhaps I will figure it out someday. Otherwise, ImageMagick remains hopeful, however I think you will have to admit, that after following all your suggestions, what I have started re: DrawRect is not as insane as it seems. You obviously have a different skillset than me at this point, so please be aware of the differences.

[jmac698]

@Dogway,
I didn't realize you had finished your function. I would like to create a draw shapes plugin, finally there will be an efficient way to do this. I need to draw some 'fancy shapes' but they aren't things that could be handdrawn, for example Avia type test charts like resolution (which involves various sinewaves).

[redfordxx]

Originally Posted by jmac698

, due to the way mt_merge works.
It calculates (x*256+128)/256 and when x=255 (when drawing a white line), I'm getting 254 back, so it's NOT drawing the correct colors.
I'll have to investigate later.

Look here try RMerge, and you are welcome to write feedback to that thread. I would appreciate it before I am adding the SSE optimization to this filter.

I had here and there also issues with this mt_merge-->254 thing, so I just btw made my own merge;-)
Tell me if this meets your needs in that case:

RMerge(c1,c2,m,mode=256,sse=0)                  #sse=0 necessary before I add optimization

[jmac698]

I needed a way to draw shapes and found my own script again.

Just to summarize the thread,

1 I investigated imagemagick and subtitles renderers; both require specifying colours only in RGB

2 Some YUV colours cannot be specified in RGB without using negative values, and neither of the above tools allow this

3 Even the range of YUV colours which can be made from positive RGB values, can't all be made, because of precision loss in the mapping between coordinate systems. (think of communicating an x,y coordinate from another coordinate system that's slanted 45 degrees; (2,2) in the slanted version maps to (1.414,1.414) in the second version, so you can see how gaps can occur).

4 In summary, there is no way to use positive, 8bit RGB colours to specify all video YUV colours.

5 I need opacity to range 0-256 because otherwise, there is an error when using all-white (white=255, 99.6% opacity=255/256, 255*255/256=254. A fully opaque white should give 255. The small error of 254 means that further masking steps won't match when looking for 255.

6 Drawing shapes in a paint program still has the same problem; you can only specify colours in RGB. Also, the shapes I need would be very tedious to create by hand - imagine drawing standard colourbars, or a frequency sweep (something which can only be calculated).

7 I'd like to expand these tools to draw sinewaves, circles and special YUV colours like I/Q (from the standard colourbars).

8 You can import raw YUV in imagemagick but it's expecting a specific raw format, which I haven't found documented. There are no options to specify variations of the format.

ps I'm actually working on an improved de-comet algorithm now, and I *still* have no better way to programaticaly create YUV test patterns

[TheFluff]

Originally Posted by jmac698

5 I need opacity to range 0-256 because otherwise, there is an error when using all-white (white=255, 99.6% opacity=255/256, 255*255/256=254. A fully opaque white should give 255. The small error of 254 means that further masking steps won't match when looking for 255.

Why on earth would you be dividing by 256? Since when does 256 fit in an unsigned 8-bit byte?

Also, here's some food for thought: RGB24 and YUV 444 both happen to have 24 bits per pixel. What a neat coincidence! If only there was an AssumePixelFormat() (hint: it's probably trivial to write). Of course, blending doesn't work the same in YUV as it does in RGB but at least it gives you a trivial way to draw shapes using existing tools that only support RGB. Overlay() should solve the rest.

[jmac698]

AssumePixelFormat would have been very useful when I was trying to do operations in YUV like http://forum.doom9.org/archive/index.php/t-108018.html. Then I could use RGB calculations and move back to YUV (but that was before I discovered masktools). I remember going through all sorts of hoops to perfectly copy YUV planes <> to RGB. I believe it was the use of YtoUV, UtoY, VtoY, and using greyscale and mergeRGB. Now I wish masktools worked in RGB! The problem remains.

I'm allowing an opacity range up to 256 for the same reason http://avisynth.nl/index.php/Overlay does, to allow fully transparent and fully opaque. 255/256 refers to the internal calculation that is done; the value of the mask is interpreted as a fraction of 256 (and yes, Overlay needs to do 16bit calculations).

You also have to be aware that Overlay does extra manipulations that I want to avoid (which is also why it's slow), such as colour interpolation.

ps anyhow, I got the job done now. I made my test video and solved a programming problem I was having. Avisynth to the rescue I made progress on my de-comet plugin, the problem is cluster analysis of Guassian mixing, and I want to use maximum expectation.

[TheFluff]

Originally Posted by jmac698

I'm allowing an opacity range up to 256 for the same reason http://avisynth.nl/index.php/Overlay does, to allow fully transparent and fully opaque.

255/255 == 1
0/255 == 0

Please explain to me in what way this does not allow fully transparent and fully opaque.

I'm not exactly sure why Overlay() uses 0-256 internally but a random guess says it's to get a right shift by 8 instead of a division by 255 in a calculation in an inner loop. In other words, it's probably a 90's optimization that doesn't really matter today, and I suspect it's actually off-by-one in some situations. It's most definitely not to allow "fully transparent and fully opaque"; any scale would allow that.

As Gavino said three years ago:

Originally Posted by Gavino

I don't see the point of having 0-256 instead of 0-255.
The latter seems more natural to me and the former introduces an unnecessary special case into the code for no appreciable gain.

Originally Posted by jmac698

You also have to be aware that Overlay does extra manipulations that I want to avoid (which is also why it's slow), such as colour interpolation.

I'm actually not at all aware of these "extra manipulations". It does upconvert to YV24 internally, but in this case we're discussing your input is already YV24, so I really don't understand what you're talking about. You could also just use mt_merge if it makes you happier, it does mostly the same thing except only in YV12. The main reasons overlay is slower are that it doesn't have a SIMD implementation and that it processes full chroma resolution.

edit: I was right about the 128 thing. mt_merge uses the same wonky 256 and like I suspected it's sometimes off-by-one; see https://github.com/tp7/masktools/issues/12

[Wilbert]

Also, here's some food for thought: RGB24 and YUV 444 both happen to have 24 bits per pixel. What a neat coincidence! If only there was an AssumePixelFormat() (hint: it's probably trivial to write). Of course, blending doesn't work the same in YUV as it does in RGB but at least it gives you a trivial way to draw shapes using existing tools that only support RGB. Overlay() should solve the rest.

You need to flip and convert to planar too, but it remains trivial

I guess the following should also work:

rgb_clip = ...
y = rgb_clip.ShowRed("Y8") # not documented yet
u = rgb_clip.ShowGreen("Y8")
v = rgb_clip.ShowBlue("Y8")
YToUV(y,u,v)

[colours]

Wilbert: actually, the U and V planes come before the Y plane for YToUV, so it should be YToUV(u,v,y) if the R/G/B planes are used to store Y/U/V information.

[StainlessS]

Creates a Marker

AVISource("E:\V\StarWars.avi")
Blk = Last.Blankclip(width=64,height=64,Length=1)
#InFix  = RT_string("(x==0|x==%d|y==0|y==%d)?128:0",Blk.Width-1,Blk.Height-1) 
InFix  = RT_string("((x==0|x==%d|y==0|y==%d)&(x<4|x>=%d)&(y<4|y>=%d))?255:0",Blk.Width-1,Blk.Height-1,Blk.Width-4,Blk.Height-4) 
Marker=Blk.mt_lutspa(relative = false,yExpr=mt_Polish(InFix), chroma = "128") 

Overlay(Marker,x=(Width-Marker.width)/2,y=(Height-Marker.Height)/2,Mode="Exclusion")

return last

sort of like a square with only corners drawn

   __     __
   |       |


   |_     _|

EDIT: Here as function to create mask for Overlay, Hit=false for some kind of 'Miss' and true for some kind of 'Hit' (detect).

Function RectMarker(clip c,int W, Int H,Bool "Hit") {
    # req RT_Stats & mt_tools_2, Returns YV12 frame WxH same FPS as c and without audio
    Hit=Default(hit,False)
    InFix = (Hit)
	\ ? RT_string("(((x<=1|x>=%d|y<=1|y>=%d)&(x<4|x>=%d)&(y<4|y>=%d)|(x==0|x==%d|y==0|y==%d)))?255:0", W-2,H-2,W-4,H-4,W-1,H-1) 
	\ : RT_string("((x<=1|x>=%d|y<=1|y>=%d)&(x<4|x>=%d)&(y<4|y>=%d))?255:0", W-2,H-2,W-4,H-4) 
    c.Blankclip(width=W,height=H,Length=1,pixel_type="YV12").Killaudio
    return mt_lutspa(relative = false,yExpr=mt_Polish(InFix), chroma = "128")
}

rough usage with some missing code

# ...

    MissMask = RectMarker(c,BlkW,BlkH,false) 
    HitMask  = RectMarker(c,BlkW,BlkH,true) 
    MissMark = MissMask.BlankClip(Color=MissColor) 
    HitMark  = HitMask.BlankClip(Color=HitColor) 

# ...

    (hit!=0) ? Overlay((hit==2)?HitMark:MissMark,x=x,y=y,Mask=(hit==2)?HitMask:MissMask,Mode="Blend") : NOP

[StainlessS]

Small update to prev post Rectal Marker, for marking your territory or something.

Function Elliptical_Marker(clip c,int W, Int H,Bool "Soft") {
    # Req mt_tools_2, Returns YV12 frame WxH same FPS as c and without audio
    Soft=Default(Soft,False)
    # INFIX_H="(((x-.5)^2 +(y-.5)^2) < .25 ? 255 : 0"                                                                       # INFIX: Elliptical Disk (Hard Edge) :: mode = "relative", Radius=0.5, Rad^2=0.25 :: SOFT_Inner rad:(Rad*0.9)^2~=0.2
    # INFIX_S="((x-.5)^2+(y-.5)^2)<.2?255:(((x-.5)^2+(y-.5)^2)<.25?(.25-((x-.5)^2+(y-.5)^2))*5100):255"                     # INFIX: Elliptical Disk (Soft Edge) :: (.25-0.2)*5100=255 : (.25-.25)*5100=0.0] :: mode="relative", 0.25=1.0 : 0.2~=0.9
    rpn = (!Soft)
    \ ? "x 0.5 - 2 ^ y 0.5 - 2 ^ + 0.25 < 255 0 ?"                                                                          [*  RPN:   Elliptical Disk [Hard Edge] *]
    \ : "x 0.5 - 2 ^ y 0.5 - 2 ^ + 0.2 < 255 x 0.5 - 2 ^ y 0.5 - 2 ^ + 0.25 < 0.25 x 0.5 - 2 ^ y 0.5 - 2 ^ + - 5100 *"      [*  RPN:   Elliptical Disk [Soft Edge] *]
    c.Blankclip(width=W,height=H,Length=1,pixel_type="YV12").Killaudio
    return Last.mt_lutspa(mode = "relative", expr = rpn, chroma = "-128" )
}

EDITED:

EDIT: And an EllipticalDisk marker, used in this case as an Overlay mask.

Function Elliptical_Marker(clip c,int W, Int H,Bool "Soft") {
    # Req mt_tools_2, Returns YV12 frame WxH same FPS as c and without audio
    Soft=Default(Soft,False)
    # INFIX_H="(((x-.5)^2 +(y-.5)^2) < .25 ? 255 : 0"                                                                       # INFIX: Elliptical Disk (Hard Edge) :: mode = "relative", Radius=0.5, Rad^2=0.25 :: SOFT_Inner rad:(Rad*0.9)^2~=0.2
    # INFIX_S="((x-.5)^2+(y-.5)^2)<.2?255:(((x-.5)^2+(y-.5)^2)<.25?(.25-((x-.5)^2+(y-.5)^2))*5100):255"                     # INFIX: Elliptical Disk (Soft Edge) :: (.25-0.2)*5100=255 : (.25-.25)*5100=0.0] :: mode="relative", 0.25=1.0 : 0.2~=0.9
    rpn = (!Soft)
    \ ? "x 0.5 - 2 ^ y 0.5 - 2 ^ + 0.25 < 255 0 ?"                                                                          [*  RPN:   Elliptical Disk [Hard Edge] *]
    \ : "x 0.5 - 2 ^ y 0.5 - 2 ^ + 0.2 < 255 x 0.5 - 2 ^ y 0.5 - 2 ^ + 0.25 < 0.25 x 0.5 - 2 ^ y 0.5 - 2 ^ + - 5100 *"      [*  RPN:   Elliptical Disk [Soft Edge] *]
    c.Blankclip(width=W,height=H,Length=1,pixel_type="YV12").Killaudio
    return Last.mt_lutspa(mode = "relative", expr = rpn, chroma = "-128" )
}


Avisource("F:\V\StarWars.avi")                                      # Source YV12
Sym=Avisource("F:\V\XMen2.avi").Trim(5000,0)                        # Overlay clip

W=(Width/16)  * 4
H=(Height/16) * 4
Sym=Sym.BilinearResize(W,H)
El = Sym.EllipticalMarker(W,H,true).Loop(Sym.Framecount)
Return OverlayClip(Last,Sym,S=100,E= -Sym.FrameCount,x= -(Sym.Width+16),y= -(Sym.Height+16),mask=El)

EDITED:

EDIT: Changed to Soft edge disk.

[StainlessS]

Script to make any movie under the sun, seem interesting. [EDIT: Req Masktools v2.0]

# BouncingBallMovie.avs

Function Elliptical_Marker(clip c,int W, Int H,Bool "Soft") {
    # Req mt_tools_2, Returns YV12 frame WxH same FPS as c and without audio
    Soft=Default(Soft,False)
    # INFIX_H="(((x-.5)^2 +(y-.5)^2) < .25 ? 255 : 0"                                                                       # INFIX: Elliptical Disk (Hard Edge) :: mode = "relative", Radius=0.5, Rad^2=0.25 :: SOFT_Inner rad:(Rad*0.9)^2~=0.2
    # INFIX_S="((x-.5)^2+(y-.5)^2)<.2?255:(((x-.5)^2+(y-.5)^2)<.25?(.25-((x-.5)^2+(y-.5)^2))*5100):255"                     # INFIX: Elliptical Disk (Soft Edge) :: (.25-0.2)*5100=255 : (.25-.25)*5100=0.0] :: mode="relative", 0.25=1.0 : 0.2~=0.9
    rpn = (!Soft)
    \ ? "x 0.5 - 2 ^ y 0.5 - 2 ^ + 0.25 < 255 0 ?"                                                                          [*  RPN:   Elliptical Disk [Hard Edge] *]
    \ : "x 0.5 - 2 ^ y 0.5 - 2 ^ + 0.2 < 255 x 0.5 - 2 ^ y 0.5 - 2 ^ + 0.25 < 0.25 x 0.5 - 2 ^ y 0.5 - 2 ^ + - 5100 *"      [*  RPN:   Elliptical Disk [Soft Edge] *]
    c.Blankclip(width=W,height=H,Length=1,pixel_type="YV12").Killaudio
    return Last.mt_lutspa(mode = "relative", expr = rpn, chroma = "-128" )
} 



Function DrawBounce(clip c) {   # Physical window coords
    Squash = (G_bounceY == 0) 
    B = (Squash) ? G_Squashed_Ball : G_Ball
    Xrel = (b.Width - G_Ball.Width)  / 2
    phyX = Round(G_bounceX * (c.Width - G_Ball.Width - xrel))
    phyY = Round((1.0 - G_bounceY) * c.Height) - B.Height
    (G_Sym.IsClip)  
        \ ? c.Overlay(Squash?G_Sym_Squashed:G_Sym,x=phyX,y=phyY,mask=B) 
        \ : c.OverLay(B.BlankClip(Color=G_BALLCOLOR,Length=1),x=phyX,y=phyY,Mask=B)    
    return Last
}

Function MoveBounce(clip c) {   # Logical coords (0.0 -> 1.0, cartesian)
    bX = G_bounceX + G_bounceXstep
    Global G_bounceXstep = (0.0 < bX < 1.0) ? G_bounceXstep : -G_bounceXstep    # Wall Bounce, Switch direction next time
    Global G_bounceX = (0.0 > bX) ? 0.0 : (bX > 1.0) ? 1.0 : bX 
    bY = Sin(G_bounceAngle) * G_bounceMax
    Global G_bounceY = (bY < 0.0) ? 0.0 : bY
    bAng = G_bounceAngle + G_bounceAngleStep
    Global G_bounceAngle = (bAng > PI) ? 0.0 : bAng 
    return c.DrawBounce
}

Function RandF() {return Rand / 32767.0}

######
#Avisource("F:\V\SexInTheCity.avi")
#Avisource("F:\V\XMen2.avi")
ColorBars.ShowFrameNumber.KillAudio

Rand(Seed=True)
###
Global G_BALLCOLOR          = $C020FF
BALLSZ                      = 128
SQUISH                      = 8 
SOFT                        = True
SYM                         = True  # False = Plain bouncing ball
###  GENERAL CONTROL for bouncer (1.0=norm) 
SPEED                       = 1.0
BOUNCEHEIGHT                = 1.0
### 
Global G_bounceMax          = BOUNCEHEIGHT * (1.0-Float(BALLSZ)/Height)
Global G_bounceXstep        = 0.005 * SPEED                  # Xpos increment
Global G_bounceAngleStep    = (0.01 + RandF * 0.04) * SPEED  # Step angle
### Start pos
Global G_bounceX            = RandF                          # Current X position
Global G_bounceY            = RandF                          # Current Y position
Global G_bounceAngle        = RandF * PI
######
Global G_Ball               = Last.Elliptical_Marker(BALLSZ,BALLSZ,SOFT)
Global G_Squashed_Ball      = Last.Elliptical_Marker(BALLSZ+SQUISH,BALLSZ-SQUISH,SOFT)
Global G_Sym                = (SYM) ? BilinearResize(BALLSZ,BALLSZ) : 0
Global G_Sym_Squashed       = (SYM) ? BilinearResize(G_Squashed_Ball.Width,G_Squashed_Ball.Height): 0
    
ScriptClip("return MoveBounce")

EDITED