VideoHelp Forum




+ Reply to Thread
Results 1 to 26 of 26
  1. hi, I have a simple script, I used the usual commands, but this time for a reason I can't get, the logo is not erased well.
    what am I doing wrong? any suggestions?

    here the sample, the encode, script and mask
    https://www.dropbox.com/sh/5p7lbb735zl71p7/AAB7G34aENbg6Ky0_KtMfjHka?dl=0

    pls help.
    Quote Quote  
  2. Originally Posted by VoodooFX View Post
    I told you that full duration sample is needed to check deblend.
    oh! ok sorry, added complete file.
    Last edited by maudit; 3rd Jan 2023 at 08:45.
    Quote Quote  
  3. Video Damager VoodooFX's Avatar
    Join Date
    Oct 2021
    Location
    At Doom9
    Search PM
    I just run it and it created good deblending masks for me, what version do you use?
    Meanwhile I'll try to reproduce your bad masks.
    Quote Quote  
  4. Originally Posted by VoodooFX View Post
    I just run it and it created good deblending masks for me, what version do you use?
    Meanwhile I'll try to reproduce your bad masks.
    I used # Version: 3.2
    now updated to 3.5

    I didn't noticed the mask was so bad.
    Now zooming into it, I see better, there are some kind of horizontal darker lines.
    when you created your good one, did you use Analyze=2?

    I'll retry as soon as possible.

    thx!
    Quote Quote  
  5. Video Damager VoodooFX's Avatar
    Join Date
    Oct 2021
    Location
    At Doom9
    Search PM
    From 3.2 there were no changes affecting this delogo job.
    What lines, again searching for microbes under microscope?

    Most likely your problem is that you analyzed frames without logo at the end. I have used "Trim(200,150938)" with Analyze=2, try 1 maybe it will be better.


    Btw, when running some tests I've encountered some reeealy weird things I can't explain... need to investigate what is going on. [This probably doesn't affect your delogo]
    Quote Quote  
  6. Video Damager VoodooFX's Avatar
    Join Date
    Oct 2021
    Location
    At Doom9
    Search PM
    Here is Show=2 from your shared masks [frame 3740 when used Trim mentioned above]:



    Here it's from my first try [frame 3740 too, but I've used a lil bit different Loc]:




    I ran few other tests to do better, but I'm still dazzled with those weird results...
    Quote Quote  
  7. Originally Posted by VoodooFX View Post
    What lines, again searching for microbes under microscope?
    ahah! microbes.. good similitude.
    I'm just noticing that the result is not so homogeneous as I usually see.


    Most likely your problem is that you analyzed frames without logo at the end. I have used "Trim(200,150938)" with Analyze=2, try 1 maybe it will be better.
    oh! ok now I see those final frames where logo disappears.
    but why (200,150938)? what's the problem with the first 200 frames?
    and also why 150938? for me the last useful frames is the 152691.
    So I think I'll try with Trim(0,152691) and I'll let you know.

    thx.
    Quote Quote  
  8. Video Damager VoodooFX's Avatar
    Join Date
    Oct 2021
    Location
    At Doom9
    Search PM
    Originally Posted by maudit View Post
    but why (200,150938)? what's the problem with the first 200 frames?
    and also why 150938? for me the last useful frames is the 152691.
    Just semi-random range, because most of the times I like to exclude intro and credits from analyze, because sometimes there are "cgi colors" that can negatively affect masks for the rest of video.
    Quote Quote  
  9. ok, i've first tried A2 with #Trim(2891,147765)
    and the result mask was just a bit better.

    then I tried A3 with and without the Trim,
    both the results are similar each other but not the same,
    and even if A3 mask is much better than A2, the deblend is still not perfect.
    some logo remnants and borders are still visible, especially in the dark background.

    how can I improve the final result?
    added both the A3 masks in the folder.
    Quote Quote  
  10. Video Damager VoodooFX's Avatar
    Join Date
    Oct 2021
    Location
    At Doom9
    Search PM
    I'll check it tomorrow [maybe].

    Btw, yesterdays "weird results" was false alarm, but today I noticed other weird stuff...

    EDIT:
    I noticed that your bmp mask file is everytime different, why you are recreating it multiple times?
    Quote Quote  
  11. Originally Posted by VoodooFX View Post
    I'll check it tomorrow [maybe].

    Btw, yesterdays "weird results" was false alarm, but today I noticed other weird stuff...

    EDIT:
    I noticed that your bmp mask file is everytime different, why you are recreating it multiple times?

    maybe it does not mean anything but there are multiple xmas animation around the logo every N minutes.
    about the bmp file, i don't really know.
    maybe I didn't fully understand how the procedure works.
    I usually delete all the files and start from scratch using those two Inpaintdelogo lines in the script.
    Quote Quote  
  12. Video Damager VoodooFX's Avatar
    Join Date
    Oct 2021
    Location
    At Doom9
    Search PM
    Looks like Analyze=3 creates best deblending masks for this video, looks pretty good for me.
    On the black frames there is hardly visible blueish remnant, you can run different parameters on them with some conditional filter.

    Originally Posted by maudit View Post
    maybe it does not mean anything but there are multiple xmas animation around the logo every N minutes.
    Maybe that's why Analyze=1 looks so bad, I'll investigate, actually I need to rewrite it.

    Originally Posted by maudit View Post
    about the bmp file, i don't really know.
    maybe I didn't fully understand how the procedure works.
    I usually delete all the files and start from scratch using those two Inpaintdelogo lines in the script.
    Basemask [bmp] you create once then no need to recreate it ten times. Different mask filename, Loc, Analyze & AnalyzeTh parameters trigger reanalysis and creates different files.
    Changing video duration usually won't trigger reanalysis, so I archive a basemask and rename it to mask_trim_100-10000.bmp for example.
    Quote Quote  
  13. Originally Posted by VoodooFX View Post
    On the black frames there is hardly visible blueish remnant, you can run different parameters on them with some conditional filter.
    ok, but which ones? exactly what params?
    what are conditional filters? I'd like to learn how to fix those blue/purple remnants that I often see in dark backgrounds, on several movies.


    Basemask [bmp] you create once then no need to recreate it ten times. Different mask filename, Loc, Analyze & AnalyzeTh parameters trigger reanalysis and creates different files.
    Changing video duration usually won't trigger reanalysis, so I archive a basemask and rename it to mask_trim_100-10000.bmp for example.
    uhm.. not sure I got it, but ok I'll try.
    Quote Quote  
  14. Video Damager VoodooFX's Avatar
    Join Date
    Oct 2021
    Location
    At Doom9
    Search PM
    Originally Posted by maudit View Post
    Originally Posted by VoodooFX View Post
    On the black frames there is hardly visible blueish remnant, you can run different parameters on them with some conditional filter.
    ok, but which ones? exactly what params? what are conditional filters?
    Like Greyscale=2 for example. Conditional filters: http://avisynth.nl/index.php/ConditionalFilter
    Probably it's too advanced for newbies, later I'll write a script for you. Better in the meantime you could familiarize with AnalyzeTh parameter.

    Btw, tested with those animations excluded, they don't influence the masks.


    Originally Posted by maudit View Post
    uhm.. not sure I got it, but ok I'll try.
    What exactly you don't understand?
    Quote Quote  
  15. Video Damager VoodooFX's Avatar
    Join Date
    Oct 2021
    Location
    At Doom9
    Search PM
    Before mangling with conditionals, maybe these settings makes it better for you:

    mmm.ebmp: https://we.tl/t-jEU52mVLmw

    Code:
    LWLibAvVideoSource("D:\In nome del Papa Re - complete.mkv")
    
    m="D:\pope king\mmm.ebmp"
    InpaintDelogo(mask=m, Loc="1638,868,-0,-0", Mode="Deblend",
    \ YAlphaSolidOuts=0, UAlphaSolidOuts=0, VAlphaSolidOuts=0,
    \ YAlphaSolidIns=99, UAlphaSolidIns=110, VAlphaSolidIns=110,
    \ YColorSolid=240, UColorSolid=128, VColorSolid=128)
    Or maybe just this:
    Code:
    LWLibAvVideoSource("D:\In nome del Papa Re - complete.mkv")
    
    m="D:\pope king\mmm.ebmp"
    InpaintDelogo(mask=m, Loc="1638,868,-0,-0", Mode="Deblend", Optimize=1)
    Quote Quote  
  16. Originally Posted by VoodooFX View Post
    Any feedback on the examples in the above post?
    https://www.dropbox.com/sh/husvgohteqhnzsn/AACCoTyO6f-9gt4QFetc1ofka?dl=0
    Quote Quote  
  17. Video Damager VoodooFX's Avatar
    Join Date
    Oct 2021
    Location
    At Doom9
    Search PM
    It's out of position because you've used Crop before delogo, there i no crop in my example scripts.
    Btw, could you post frame numbers where "some logo remnants and borders are still visible"?
    Quote Quote  
  18. Originally Posted by VoodooFX View Post
    It's out of position because you've used Crop before delogo, there i no crop in my example scripts.
    Btw, could you post frame numbers where "some logo remnants and borders are still visible"?
    sry for the delay, ok I got it, removed the crop, now it works, anyway the result is similar to the script I used.
    logo blue/purple remnants are almost everywhere on black backgrounds, for example the frames: 7483 or 8380 or others.

    That "Optimize=1" is producing bad results like this:
    https://www.dropbox.com/sh/um5whihzhg7fcdm/AADhUk1M9EpJWDvfbxyF1mYLa?dl=0
    "Optimize=2" is better there, but not on many other frames.

    I'm noticing that adding params like these: "Interp=4, IntSpd=4, Optimize=2" helps a lot in dark background but also produce light distortion effects on other frames, expecially on moving scenes, and also some excessive blurry effects on colored frames.

    So I think I'll use my mask "A3 Trim(2891,147765)", with GrainPP=1 param only, forgetting about remnants on dark frames.
    Quote Quote  
  19. Video Damager VoodooFX's Avatar
    Join Date
    Oct 2021
    Location
    At Doom9
    Search PM
    At 7483 I fail to see any remnants, at 8380 there is some visible.
    Maybe I could implement a conditional internally to switch to "Interp=4" automatically when frames have solid color look.

    Btw, "Optimize=2" auto disables Interp on higher contrast frames, for example frames like your mentioned 7483.
    Quote Quote  
  20. Originally Posted by VoodooFX View Post
    Maybe I could implement a conditional internally to switch to "Interp=4" automatically when frames have solid color look.
    cool, it's a good idea. thx!
    Quote Quote  
  21. Video Damager VoodooFX's Avatar
    Join Date
    Oct 2021
    Location
    At Doom9
    Search PM
    Implemented in v3.6, check if it makes a difference [it's enabled by default].

    Later I'll check if conditional for ~greyscale frames can be implemented too, this should improve things further.
    Quote Quote  
  22. Originally Posted by VoodooFX View Post
    Implemented in v3.6, check if it makes a difference [it's enabled by default].

    Later I'll check if conditional for ~greyscale frames can be implemented too, this should improve things further.

    I will no longer publish this movie cause apart those xmas animation around the logo, I've found another intrusive AD animated logo at 42300
    and probably elsewhere.
    Anyway I continue to use it as a test, I've upgraded to 3.6, it says:

    "Optimize" is split to "Optimize1" & "Optimize2"."

    what it means? how it works now?
    avspmod no longer recognize Optimize=1.

    this ver 3.6 looks like pretty better on dark scene, but I still can see blue/purple borders somewhere, on hi-contrast frames I think.
    so maybe I can add some parameter. idk..
    Quote Quote  
  23. Video Damager VoodooFX's Avatar
    Join Date
    Oct 2021
    Location
    At Doom9
    Search PM
    Originally Posted by maudit View Post
    I've upgraded to 3.6, it says:

    "Optimize" is split to "Optimize1" & "Optimize2"."

    what it means? how it works now?
    avspmod no longer recognize Optimize=1.
    That means that there is no "Optimize" anymore, there are "Optimize1" & "Optimize2" instead.
    Everything is in the manual:

    Code:
    # Int "Optimize1"   :  Enables conditional inpainting optimizations (from 0 to 1; default: 1).
    #                     Not made to work with the dynamic masks (likely there wouldn't be a lot of benefit).
    #                     1 -  Optimizations for the slow "Turbo" & "IntSpd" presets [up to ~1000% speed-up].
    #                         Enables auto switching to the faster presets on the frames where logo is in ~solid background.
    
    
    # Int "Optimize2"   :  Enables conditional optimizations for "Interp" (from 0 to 3; default: 3).
    #                     1 -  Turns off "Interp" for frames where logo is in the background with higher contrast.
    #                     2 -  Switch to max "Interp" for frames where logo is in ~solid background.
    #                     3 -  Both 1 & 2 are enabled.

    Originally Posted by maudit View Post
    this ver 3.6 looks like pretty better on dark scene, but I still can see blue/purple borders somewhere, on hi-contrast frames I think.
    so maybe I can add some parameter. idk..
    Because on higher contrast background "Interp" is disabled by default [aka Optimize2=1], you can turn it off by "Optimize2=0" or "Optimize2=2".
    But then there is some possible "Interp" artifacts, then use "IntSpd=4" to reduce them.

    Probably you don't understand what that "Interp" does, so here you go:
    "Interp" controls how much data is merged from the inpainted clip to the deblended clip, higher "Interp" = more data is merged.
    You can see an inpainted clip in Show=2 [indicated by "IntSpd"].
    Quote Quote  



Similar Threads

Visit our sponsor! Try DVDFab and backup Blu-rays!