New Plugin: Decomet 0.4

[StainlessS]

Post #34 modified.

@Jmac,

I was curious to know what values of U and V are valid at both Y=16 and Y=235, so

GSCript("""
    Z=5    Got=0
    RT_DebugF("Lo=# Y=$10(16)")
    for(ui=128-Z,128+Z) {
        for(vi=128-Z,128+Z) {            
            q=(16*256 + ui)*256+vi
            c=BlankClip(pixel_type="YV12",color_yuv=q)
            c2=c.ConvertToRGB32
            current_frame=0
            R2=Round(c2.ShowRed(pixel_Type="YV12").AverageLuma)
            G2=Round(c2.ShowGreen(pixel_Type="YV12").AverageLuma)
            B2=Round(c2.ShowBlue(pixel_Type="YV12").AverageLuma)
            c3=c2.ConvertToYV12
            Y3=Round(c3.AverageLuma)
            U3=Round(c3.AverageChromaU)
            V3=Round(c3.AverageChromaV)
            if(U3==ui && V3==vi && Y3==16) {
                GOT=GOT+1
                RT_debugF("%d] YUV = $%0.2X%0.2X%0.2X (RGB = $%0.2X%0.2X%0.2X)",GOT,Y3,U3,V3,R2,G2,B2)
            }
        }
    }
    Got2=0
    RT_DebugF("Hi=# Y=$EB(235)")
    for(ui=128-Z,128+Z) {
        for(vi=128-Z,128+Z) {            
            q=(235*256 + ui)*256+vi
            c=BlankClip(pixel_type="YV12",color_yuv=q)
            c2=c.ConvertToRGB32
            current_frame=0
            R2=Round(c2.ShowRed(pixel_Type="YV12").AverageLuma)
            G2=Round(c2.ShowGreen(pixel_Type="YV12").AverageLuma)
            B2=Round(c2.ShowBlue(pixel_Type="YV12").AverageLuma)
            c3=c2.ConvertToYV12
            Y3=Round(c3.AverageLuma)
            U3=Round(c3.AverageChromaU)
            V3=Round(c3.AverageChromaV)
            if(U3==ui && V3==vi && Y3==235) {
                GOT2=GOT2+1
                RT_debugF("%d] YUV = $%0.2X%0.2X%0.2X (RGB = $%0.2X%0.2X%0.2X)",GOT,Y3,U3,V3,R2,G2,B2)
            }
        }
    }
""")

return Messageclip("Done")

00000003    1.61430311  [3680] RT_DebugF: Lo=# Y=$10(16)    
00000004    3.09199762  [3680] RT_DebugF: 1] YUV = $108080 (RGB = $000000)  
00000005    3.36213446  [3680] RT_DebugF: 2] YUV = $108180 (RGB = $000002)  
00000006    3.64207911  [3680] RT_DebugF: 3] YUV = $108280 (RGB = $000004)  
00000007    4.60943556  [3680] RT_DebugF: Hi=# Y=$EB(235)   
00000008    5.66757488  [3680] RT_DebugF: 3] YUV = $EB7E80 (RGB = $FFFFFB)  
00000009    5.98870993  [3680] RT_DebugF: 3] YUV = $EB7F80 (RGB = $FFFFFD)  
00000010    6.32429600  [3680] RT_DebugF: 3] YUV = $EB8080 (RGB = $FFFFFF)

Ooops, bug fix (added compare on Y).

[jmac698]

Hi,
Thanks for ressurecting this to source.

However, I think you misunderstood a number of things about comets. Not your fault, as you haven't really worked with them. For one thing, they are not really like the demo at all. They aren't necessarily white, and they fade along the streak. The black lines aren't the same as comets really. They are more like a shadow from strong luma transitions, and not always present with the white comets. It was just that I could handle those at the same time. Finally, you should never average the frames, that just makes it worse. Also there's never meant to be two luma points. The idea was to bring the video towards the most middle value, so there's one point. Also the colour isn't really affected by the comets or the black shadows.

I will have to fix up the source. The laser disc sample might give me ideas on how to improve it, but for now I would pre-correct the brightness so the two copies look similar.

[StainlessS]

OK JM,
Its all in your hands, the SuperBlack/White thing will I hope be of use in deciding what to do.
I'll bow out now.

Toodle-ooooo.

[Busty]

@ jmac698: if I can help out with more samples or a short clip let me know what you need and I'll gladly provide them.

[StainlessS]

@Busty, I see that you images from post #39 have yet to be approved.

[jmac698]

@busty
Yes I'll work on this in about a week, I have some ideas to make decomet better and also single frame mode. So find a fairly heavy sample or a spot where there's dozens over the clip length, so I get a lot of examples to profile. 3 recordings if possible so I can compare the true median cleaned to my plugin. Thanks

Link to DePulse in case anyone is looking.
https://forum.doom9.org/showthread.p...et#post1800999

[Busty]

@jmac698:
I can only provide two captures of the same video in this case, as I only have two copies of that laserdisc. That's also the reason why I was hoping decomet could help, otherwise I'd have checked out toot or median. I did use median on another occasion, but it didn't have that much problems to begin with.

If I come across a problematic part where I have three captures I'll upload a sample.

I extracted a ten second portion of a problematic part of said laserdisc, which results in 2 x 74 MB (utvideo interlaced PAL 768x576). If that size is too much I'll gladly provide a shorter sample, just let me know.

This is the link for the 2 files, expires on August 11.: https://we.tl/LlA4hcTZGO

@ StainlessS:
can I do something about those screencaps? Btw, I checked, and I did upload in png format. Strange that they show up as jpg.

[StainlessS]

Originally Posted by Busty

TZGO[/url]
can I do something about those screencaps? Btw, I checked, and I did upload in png format. Strange that they show up as jpg.

Give time for JMac to take a peek, him da man.

[jmac698]

@busty
np, I forgot we were talking about laserdiscs here. The size is not a problem.

I've always felt we need some kind of generic video matching, that lines up temporally, matches colours, and matches geometry, all while being robust to comets and noise