VideoHelp Forum

Try DVDFab and copy Ultra HD Blu-rays and DVDs! Or rip iTunes movies and music! Download free trial !
+ Reply to Thread
Page 1 of 2
1 2 LastLast
Results 1 to 30 of 33
Thread

Threaded View

  1. I captured some video with a MiniDV camera using the camera "widescreen" function. (It was a bad idea, let's not discuss that).

    So now I have a 720x480 AVI-DV file with black bars.

    I created a DVD with Adobe Encore and flagged the video as 16:9. However when displayed on a Widescreen TV (16:9) it still have black bars at the bottom and top. I want to display it with NO black bars on a widescreen TV.

    Here is a example of the video http://rapidshare.com/files/23284654/wide.avi
    (Keep in mind that the original video is an interlaced AVI-DV, the example is de-interlaced and divx encoded)

    As you can see if you crop the video it would result in a 720x360 file, however it should be something like 720x405 to be WS as far as I know.

    Should I crop to 720x360, resize to 720x405 and then letterbox to 720x480 or what? If I do that, wouldn't it make problems with interlacing?

    Any help would be appreciated...I searched the forum, but couldn't find any answer.
    Quote Quote  
  2. Always Watching guns1inger's Avatar
    Join Date
    Apr 2004
    Location
    Miskatonic U
    Search Comp PM
    What you need to do is crop it to 720 x 360, the resize it to 720 x 480, then encode it as 16:9, author it as 16:9, and it will play back correctly.

    However, you need to be careful doing the resize because you source is interlaced, and if you resize incorrectly, you will get mud.

    The simplest solution, IMO, for a quality crop/resize, is avisynth with LeakKernalDeint. If you search the editing and restoration forums you will find examples. I'll see what i can dig up and will post back here.
    Read my blog here.
    Quote Quote  
  3. Член BJ_M's Avatar
    Join Date
    Jul 2002
    Location
    Canada
    Search Comp PM
    no need to deinterlace dv source material ..

    crop and resize to 720x480 - flag as anamorphic
    "Each problem that I solved became a rule which served afterwards to solve other problems." - Rene Descartes (1596-1650)
    Quote Quote  
  4. Always Watching guns1inger's Avatar
    Join Date
    Apr 2004
    Location
    Miskatonic U
    Search Comp PM
    So it is OK to resize up interlaced material ?

    Isn't this contrary to most advice given here ?
    Read my blog here.
    Quote Quote  
  5. Член BJ_M's Avatar
    Join Date
    Jul 2002
    Location
    Canada
    Search Comp PM
    depends how it is done -- if using virtualdub or avisynth , of course it is best to use avisynth and split fields .. if using vegas or shake or fusion or AE or something -- not necessary really ...

    LOTS of guides and instructions here on methods ...

    my point on conversion of DV to progressive - is that in most cases it is counter productive to final end quality .... unless you are going for a certain "look" or something ....
    "Each problem that I solved became a rule which served afterwards to solve other problems." - Rene Descartes (1596-1650)
    Quote Quote  
  6. Originally Posted by sirblackjack
    As you can see if you crop the video it would result in a 720x360 file, however it should be something like 720x405 to be WS as far as I know.
    No, DV does not use square pixels (note that 720/480 is not equal to 4/3). So you cannot simply take 720, multiply by 9 and divide by 16 to get 405. You have to take into account the non-square pixels. The correct procedure is to crop down to 720x360 and resize what's left to 720x480.

    As you surmised, you need to be careful about resizing interlaced material. You have to use an interlace-aware resize filter (BJ_M listed several apss that have this) or separate the fields, resize, and weave the fields back together. The best resizing methods I've seen are based around AVISynth's LeakKernelBob() filter.

    Since you're using VirtualDub -- its interlace option on the resize dialog isn't the greatest.
    Quote Quote  
  7. Thanks a lot for the help! I'm a newbie in video, but after searching the forum I found this.

    I downloaded the "KernelDeint" plugin for Avisynth, created and .avs and used the following code (I modified just some numbers)

    AVISource("movie.avi") -> No comments
    LeakKernelBob(Order=0) -> No idea what it does, but I guess order=0 means bottom fields first?
    Crop (0,60,0,-60) -> Crop the black bars (60 pixels each)
    LanczosResize(720, 480) -> Resizes to 720x480
    SeparateFields() -> Separate the fields (de-interlace?)
    SelectEvery(4,0,3) -> Select some of the fields?
    Weave() -> Put back the fields in frames (interlace?)

    Anyway I tried the code. Opened in Virtualdub, saved the file as 16:9 AVI-DV file and it worked just fine. The file now is without the black bars and interlaced.

    One final newbie question that I haven't found the answer yet. If you open an interlaced AVI-DV file in WMM, add some text/photos, then save it again as AVI-DV file. It looks like a progressive file now. So, I guess if editing is better to de-interlace first if editing the video or not?

    Thanks a lot for all the help!
    Quote Quote  
  8. Member thecoalman's Avatar
    Join Date
    Feb 2004
    Location
    Pennsylvania
    Search PM
    Originally Posted by sirblackjack
    So, I guess if editing is better to de-interlace first if editing the video or not?
    Never deinterlace footage you intend to view on a TV. For that matter you can apply that to video you want to watch on your computer too, most player do a very good job of deinterlacing on the fly. I'm not that familiar with WMM but it shouldn't be doing that, if it is I wouldn't use it.

    I captured some video with a MiniDV camera using the camera "widescreen" function. (It was a bad idea, let's not discuss that).
    Why not discuss it? this is a discussion board. :P Some cams have that feature, for what I don't know because it's useless. In case you don't know you should be able to create full frame wide screen. Look in the manual or see if there is anther setting for anamorphic wide screen, it should have it.
    Quote Quote  
  9. Member Alex_ander's Avatar
    Join Date
    Oct 2006
    Location
    Russian Federation
    Search Comp PM
    The script you used is about correct but in case of BFF (it's important to know it exactly) there should be other numbers (sorry, can't remember them now) for selecting fields. And yes, not all frames are used in the new video since bobbing doubles framerate (each field after bob represents a progressive frame with full number of lines). Other questions in your script comments: true.
    Quote Quote  
  10. Originally Posted by sirblackjack
    AVISource("movie.avi") -> No comments
    LeakKernelBob(Order=0) -> No idea what it does, but I guess order=0 means bottom fields first?
    Crop (0,60,0,-60) -> Crop the black bars (60 pixels each)
    LanczosResize(720, 480) -> Resizes to 720x480
    SeparateFields() -> Separate the fields (de-interlace?)
    SelectEvery(4,0,3) -> Select some of the fields?
    Weave() -> Put back the fields in frames (interlace?)
    LeakKernelBob() creates a full frame from each field of your source. It does so with very little artifacting. You now have twice as many frames as in the original file. Yes, "order=0" means bottom field first.

    Separate fields now splits each of those frames into top and bottom fields. Each original frame of video is now 4 fields.

    SelectEvery(4,0,3) means: for every 4 frames (which are really now fields) keep frames 0 and 3, throw away 1 and 2. You now have two fields from each frame of the original. Assuming bottom field first (the default in AVIsynth and DV AVI) #0 is a bottom field, #3 is a top field.

    Weave() then takes the two remaining fields and weaves them back together into interlaced frames. You now have one frame of output from each frame of the input. The resulting field order is the assumed BFF.
    Quote Quote  
  11. Member edDV's Avatar
    Join Date
    Mar 2004
    Location
    Northern California, USA
    Search Comp PM
    I'm confused by this thread. Which MiniDV camcorders in widescreen mode produce letterbox**? All mine (mostly Sony's) produce horizonally compressed (full vertical) 480i.

    ** the video will show as letterbox in a 4:3 viewfinder.
    Recommends: Kiva.org - Loans that change lives.
    http://www.kiva.org/about
    Quote Quote  
  12. Member Alex_ander's Avatar
    Join Date
    Oct 2006
    Location
    Russian Federation
    Search Comp PM
    Jagabo, I mentioned different numbers just because have seen it (different conditions for TFF and BFF) today in some template. From your explanation I tend to think that if selected fields were (4,1,2) this would simply change fields order (BFF to TFF here or TFF to BFF in another case). Is it true?
    Quote Quote  
  13. Member edDV's Avatar
    Join Date
    Mar 2004
    Location
    Northern California, USA
    Search Comp PM
    OK, here is my understanding. Is this wrong?

    Source is a simple DCR-TRV103 in wide mode (first and lowest Digital8 model) pointed at a circle. All my other D8/MiniDV camcorders produce this same result.

    Here is what it looks like in 720x480 over IEEE-1394 as seen in WinDV monitor.


    We all know that 720x480 pixels are not square. So here is the 640x480 square 4:3 version (H adjusted in Photoshop) which appears full height and horizontally squeezed.


    Here is the 853x480 square pixel 16:9 version again adjusted in Photoshop. The knob is round.


    This matches exactly what I see in a Vegas monitor in DV wide project settings and when monitoring it fills a 16:9 TV screen.

    PS: to see the wider image you need to double click the image in this web software to see the full width resolution.
    Quote Quote  
  14. Member thecoalman's Avatar
    Join Date
    Feb 2004
    Location
    Pennsylvania
    Search PM
    Originally Posted by edDV
    I'm confused by this thread. Which MiniDV camcorders in widescreen mode produce letterbox**?
    I don't know but they are out there, I saw another post about this before,
    Quote Quote  
  15. Always Watching guns1inger's Avatar
    Join Date
    Apr 2004
    Location
    Miskatonic U
    Search Comp PM
    Quite a few older DV cameras did letterbox instead of pseudo-anamorphic 16:9. I know a number of early Panasonic DV cameras do this.
    Read my blog here.
    Quote Quote  
  16. Member edDV's Avatar
    Join Date
    Mar 2004
    Location
    Northern California, USA
    Search Comp PM
    OK
    Recommends: Kiva.org - Loans that change lives.
    http://www.kiva.org/about
    Quote Quote  
  17. Member edDV's Avatar
    Join Date
    Mar 2004
    Location
    Northern California, USA
    Search Comp PM
    To complete the Sony DV 16:9 wide story, the camcorder 4:3 LCD shows letterbox 16:9 while the viewfinder, analog and digital outputs show full vertical, horizontally compressed 480i like the first picture above.

    Quote Quote  
  18. Member Alex_ander's Avatar
    Join Date
    Oct 2006
    Location
    Russian Federation
    Search Comp PM
    Originally Posted by Alex_ander
    Jagabo, I mentioned different numbers just because have seen it (different conditions for TFF and BFF) today in some template. From your explanation I tend to think that if selected fields were (4,1,2) this would simply change fields order (BFF to TFF here or TFF to BFF in another case). Is it true?
    Now I understand the answer is NO. Just found that template script with (4,1,2) and analyzed it. Here is the main part of it:

    LeakKernelBob(Order=1)#if TFF, Order=0 if BFF
    #Crop, Resize, AddBorders, or other filtering here
    AssumeTFF()# if TFF, Assume BFF() if BFF
    SeparateFields()
    SelectEvery(4,0,3)#if TFF, SelectEvery(4,1,2) if BFF
    Weave()


    and it is incorrect: if for separating fields the assumed fields order corresponds to the source, (4,0,3) should be used in both TFF and BFF cases. This returns original interlaced video in the end if no filtering is applied after bob. If reverse order is assumed, it's the only case where (4,1,2) works correctly, otherwise the output video would include rebuilt fields only (fields with interpolation of motion parts).
    So the script sirblackjack uses is correct.
    Quote Quote  
  19. Member
    Join Date
    Aug 2004
    Location
    United States
    Search Comp PM
    I looked at the deinterlaced sample...where do you get 720x360? Cropping borders comes out to around 720x380.
    Quote Quote  
  20. Always Watching guns1inger's Avatar
    Join Date
    Apr 2004
    Location
    Miskatonic U
    Search Comp PM
    You have to remember that we are working with non-square pixels.

    16:9 = 16/9 = 1.7778

    720 x 360 = 1.7778 : 1 = 16:9

    720 x 380 = 1.6842 : 1

    Resizing from 360 x 480 and encoding as 16:9 will give non-distorted results. Resizing from 380 to 480 will give incorrectly distorted results.
    Read my blog here.
    Quote Quote  
  21. Member
    Join Date
    Aug 2004
    Location
    United States
    Search Comp PM
    Yes...but doesn't he want to crop just the black? If he crops 60 top, 60 bottom....he will crop parts of the picture. From the sample there really was no object to guage a correct ar when converting to 16:9, except a shadowy figure at the end.
    Quote Quote  
  22. Always Watching guns1inger's Avatar
    Join Date
    Apr 2004
    Location
    Miskatonic U
    Search Comp PM
    What he wants is to watch this on a widescreen TV with no black bars.

    If you want to preserve the correct aspect ratio, then 120 lines have to go. They do not necessarily have to be split 60 top and 60 bottom, although if you want to do it fast, that is the simplest way. Having done this a couple of times from footage taken with 4:3 cameras, my preference is to do the letterbox framing/cropping manually so I can move the image within the frame as required to maintain focus on the important action.

    You have to remember as well that what you have seen is a sample out of context. What else is in the footage you have not seen ?
    Read my blog here.
    Quote Quote  
  23. Member
    Join Date
    Aug 2004
    Location
    United States
    Search Comp PM
    Originally Posted by guns1inger
    If you want to preserve the correct aspect ratio, then 120 lines have to go. They do not necessarily have to be split 60 top and 60 bottom, although if you want to do it fast, that is the simplest way.
    Yes you're right it's the fastest and simpler way to preserve the ar and convert to 16:9, although it doesn't keep the whole picture. In my mind was thinking of cropping away black (keeping the whole pic) then adding borders to the sides.
    Quote Quote  
  24. Originally Posted by Alex_ander
    Originally Posted by Alex_ander
    Jagabo, I mentioned different numbers just because have seen it (different conditions for TFF and BFF) today in some template. From your explanation I tend to think that if selected fields were (4,1,2) this would simply change fields order (BFF to TFF here or TFF to BFF in another case). Is it true?
    Now I understand the answer is NO. Just found that template script with (4,1,2) and analyzed it. Here is the main part of it:

    LeakKernelBob(Order=1)#if TFF, Order=0 if BFF
    #Crop, Resize, AddBorders, or other filtering here
    AssumeTFF()# if TFF, Assume BFF() if BFF
    SeparateFields()
    SelectEvery(4,0,3)#if TFF, SelectEvery(4,1,2) if BFF
    Weave()


    and it is incorrect:
    No, it's not incorrect, if you understand that the aim is to return TFF. That sounds like something I may have written, and while I have no desire to track it down, I usually finish by saying to encode for TFF. If you really do want to encode the same as the source (what difference does it make?), then you can do as you suggested. However, DVD is usually TFF (not always, of course), and as I always use CCE, it requires no adjustment of the settings when opening a script like that in CCE.

    sirblackjack's script, the one jagabo was commenting on, started with and returned BFF. In such cases, it should also be encoded as BFF. sirblackjack's script was correct and jagabo's comments on the script were correct. In fact, I don't see anything for you to disagree with or to call wrong.
    Quote Quote  
  25. Member Alex_ander's Avatar
    Join Date
    Oct 2006
    Location
    Russian Federation
    Search Comp PM
    Manono, I meant the following. Everything is correct here for TFF.
    But what happens in case BFF input/bob(order=0) and when the fields in bob frames are separated assuming BFF (or default, also BFF), and (1,2) numbers instead of (0,3) of the separated fields are used (according to the comments in the script)? First I thought this would turn BFF to TFF but it's not that simple. Let's imagine how the original fields are represented in bob frame sequence (the two consequent frames built from a single input frame):
    Frame1
    Top field - rebuilt (motion parts deinterlaced)
    Bottom field - original
    Frame2
    Top field - original
    Bottom field - rebuilt
    Let's first assume BFF (as in script comment) and separate fields, we get from the above (with no filtering after bob):
    0 - Bottom field - original
    1 - Top field - rebuilt
    2 - Bottom field - rebuilt
    3 - Top field - original
    Now if we take 0,3 it's OK for BFF output as Jagabo explained. We can output 1,2 for TFF and it will work somehow but both fields in all output frames will be altered without any filter applied.
    Let's now assume TFF for separating fields for BFF source (also bobbed with order=0):
    0 - Top field - rebuilt
    1 - Bottom field - original
    2 - Top field - original
    3 - Bottom field - rebuilt
    Here we must use (1,2) to keep original fields and we still get BFF but this time we didn't follow the comment 'if BFF' from the script. For outputting TFF for BFF source here the selected numbers can be reversed to (2,1).
    Do I misunderstand something? Maybe bob function of LKD changes fields order for BFF before doing its job?
    Quote Quote  
  26. AVISynth goes to great extents to maintain field order. It keeps track of which fields are top and bottom. When it weaves them back together it will maintain the field order. SelectEvery(4,1,2) indeed reverses the order in which the fields are now pipelined, but I believe AVISynth will throw away the first field to thwart your attempt to reverse field order.

    Normal procedure with SelectEvery(4,0,3)
    Code:
    1 2 3 4 -- (frames after LeakKernelBob())
    1t 1b 2t 2b 3t 3b 4t 4b -- (after AssumeTFF() SeperateFields())
    1t 2b 3t 4b -- (after SelectEvery(4,0,3)
    [1t+2b] [3t+4b] (after Weave(), TFF maintained)
    Modified procedure with SelectEvery(4,1,2)
    Code:
    1 2 3 4 -- (frames after LeakKernelBob())
    1t 1b 2t 2b 3t 3b 4t 4b -- (after AssumeTFF() SeperateFields())
    1b 2t 3b 4t -- (after SelectEvery(4,1,2)
    [2t+3b] [4t+5b] (after Weave(), TFF maintained)
    Quote Quote  
  27. Never deinterlace footage you intend to view on a TV. For that matter you can apply that to video you want to watch on your computer too, most player do a very good job of deinterlacing on the fly. I'm not that familiar with WMM but it shouldn't be doing that, if it is I wouldn't use it.
    I checked the video again. It looks like the original video is still interlaced, but the added text and photos are progressive.
    If the TV and DVD player suppports progressive scan, it should not matter if the footage is interlaced or progressive or not? (Well I asume there must be some quality lost if you deinterlace)

    Why not discuss it? this is a discussion board. Some cams have that feature, for what I don't know because it's useless. In case you don't know you should be able to create full frame wide screen. Look in the manual or see if there is anther setting for anamorphic wide screen, it should have it.
    It doesn't have the anamorphic option. The camcorder 4:3 LCD shows letterbox 16:9 while using the viewfinder (as edDV shows), but also outputs letterbox (it just adds black bars).

    Quite a few older DV cameras did letterbox instead of pseudo-anamorphic 16:9. I know a number of early Panasonic DV cameras do this.
    Indeed. The camera is an old Panasonic

    LeakKernelBob() creates a full frame from each field of your source. It does so with very little artifacting. You now have twice as many frames as in the original file. Yes, "order=0" means bottom field first.

    Separate fields now splits each of those frames into top and bottom fields. Each original frame of video is now 4 fields.

    SelectEvery(4,0,3) means: for every 4 frames (which are really now fields) keep frames 0 and 3, throw away 1 and 2. You now have two fields from each frame of the original. Assuming bottom field first (the default in AVIsynth and DV AVI) #0 is a bottom field, #3 is a top field.

    Weave() then takes the two remaining fields and weaves them back together into interlaced frames. You now have one frame of output from each frame of the input. The resulting field order is the assumed BFF.
    Thanks for the explanation
    Quote Quote  
  28. Originally Posted by Alex_ander
    Manono, I meant the following. Everything is correct here for TFF.
    But what happens in case BFF input/bob(order=0) and when the fields in bob frames are separated assuming BFF (or default, also BFF), and (1,2) numbers instead of (0,3) of the separated fields are used (according to the comments in the script)? First I thought this would turn BFF to TFF but it's not that simple.
    I don't know. I did some testing. I had a TFF interlaced video. I made it BFF with this:

    SeparateFields().Trim(1,0)
    Weave()

    BFF now. To make sure I put this after it:

    AssumeBFF().SeparateFields()

    and it played smoothly. I also checked by adding Info() to the end of the script, and it showed it to be BFF. I removed AssumeBFF().SeparateFields() and used this script on the now BFF video:

    LeakKernelBob(Order=0)
    SeparateFields()
    AssumeBFF()
    SelectEvery(4,1,2)
    Weave()

    From what I said earlier, it reversed the field order and became TFF. You said it didn't, or maybe you said doing this gives inferior results. I'm not really sure why you brought it up when it wasn't used anywhere in this thread before you mentioned it. How do you test? I tested 4 ways. I put on:

    AssumeTFF().SeparateFields()

    and it played smoothly. When I put on:

    AssumeBFF().SeparateFields()

    it played jerky. In addition, I added:

    Info()

    to the bottom of the script, and it showed it to be "Parity: Top Field First". I encoded it for TFF in CCE, authored it for DVD using Muxman, burned it to CD-RW as a Mini DVD, and tested it in my player where it played smoothly. I don't know what else to do.

    I think one of your arguments is that it's better to keep to the original field order because you'll be getting back the unaltered original fields if you're not filtering. But the only reason to do something like this is if you intend to do some filtering of an interlaced source, in this case some cropping and resizing to convert from 4:3 to 16:9, so the original fields are going to disappear during the resize.
    Quote Quote  
  29. Member Alex_ander's Avatar
    Join Date
    Oct 2006
    Location
    Russian Federation
    Search Comp PM
    Originally Posted by manono
    From what I said earlier, it reversed the field order and became TFF. You said it didn't. At least that's what I think you said.
    I didn't mean selecting (1,2) wouldn't turn BFF to TFF, of course it will.
    Quoting myself:
    Let's first assume BFF (as in script comment) and separate fields, we get from the above (with no filtering after bob):
    0 - Bottom field - original
    1 - Top field - rebuilt
    2 - Bottom field - rebuilt
    3 - Top field - original
    ... We can output 1,2 for TFF and it will work somehow but both fields in all output frames will be altered without any filter applied.


    All I tried to say was that if you use this process for turning BFF to TFF as well, you'll have to throw away original fields and use rebuilt fields only. It (at least) looks to be a 'cleaner' way to use 'original' fields (I understand they will be modified depending on ammount of filtering applied) by selecting (4,0,3) and change to TFF for encoding in the end of the script.
    Quote Quote  
  30. Originally Posted by Alex_ander
    I didn't mean selecting (1,2) wouldn't turn BFF to TFF, of course it will.
    Well, it looks like that is correct. I swear I tried this long ago and it wouldn't reverse field order. In fact, the AVISynth documentation states:

    All AviSynth filters keep track of field parity, so Weave will always join the fields together in the proper order. If you want the other order, you'll have to use ComplementParity, AssumeTFF or AssumeBFF beforehand or SwapFields afterwards.
    But I just ran the experiment and SelectEvery(4,1,2) did reverse the field order.
    Quote Quote  



Similar Threads