|
|
|
DVD-RB And 29.97fps Interlaced
|
|
Staff Member
2 product reviews
|
6. August 2004 @ 12:33 |
Link to this message
|
This was posted by happyuser following a "heated" discussion about my DVD-RB guide. I've copied it to this new thread because it's really beyond the scope of the thread it was in and is, in fact, a more advanced topic.
Quote:
Here is the questions: actually two very relevant questions for all of us:
1. You have a dvd of pure NTSC ( 29.976 ) interlaced. Are you willing to encode this dvd with DVD-Rebuilder while knowing that this DVD-Rebuilder will use "Forced Film" on this 29.976 pure interlaced to bring it down to 23.976 and then this DVD-Rebuilder will encode this pure interlaced content as interlaced ( alternate_scan=true )?
If you DISABLE "INTERLACED" for all VTS-Sets, this DVD-Rebuilder will encode every VTS-Sets as Progressive without deinterlacing anything.
2. You have a dvd that the main film ( movie ) is of Progressive ( 97% or higher FILM ( Progressive ): one that we can use "forced film" feature on ): obviously you have to use the DVD-Rebuilder to encode this DVD: because DVD-Rebuilder handles well contents of ( Progressive ) film well.
But EXTRAS ( trailers, behind the scenes, etc. ) ( on this same dvd ): some are FILM, some are contents that need to be IVTCed, and some are PURE interlaced ( NTSC ). Obviously, FILM and contents that need to be IVTCed are handled well by this DVD-Rebuilder. But PURE interlaced NTSC contents will not handle well in this manner as follows:
a. Let's say, for example, VTS_05 has six ( 6 ) video segments ( AVSs ) of either FILM or contents that need to be IVTCed, and this VTS_05 also has twenty ( 20 ) video segments ( AVSs ) that are pure interlaced ( NTSC ). Obviously, if you DO NOT disable "Interlaced" for this VTS_05 and use "Deinterlace with DECOMB", DVD-Rebuilder will encode the above twenty ( 20 ) pure interlaced video segments ( AVSs ) as interlaced ( alternate_scan=true ) and with the lines "FieldDeinterlace(blend=false)" and the "ConvertToYUY2(interlaced=true)". What would be an issue is that FieldDeinterlacer is a bad deinterlacer, and ConvertToYUY2(interlaced=true) will result in some form of color shifting. Further more, interlaced encoding is far worse than Progressive encoding.
b. So now you will not use "Deinterlace with DECOMB": DVD-Rebuilder will not deinterlace the above twenty pure interlaced video segments ( AVSs ) and encode them as interlaced ( alternate_scan=true ) and with ConvertToYUY2(interlaced=true) at 23.976 fps. According to documentation of DVD-Rebuilder, ConvertToYUY2(interlaced=true) will result in some form of color shifting.
c. Now you will DISABLE "INTERLACED" for the above VTS_05: DVD-Rebuilder will encode all the video segments ( AVSs ) as Progressive ( progressive=true ). Now you go into the twenty pure interlaced video segments ( AVSs ) above and insert statement to deinterlace with SmoothDeinterlacer into every single one of these twenty pure interlaced video segments ( AVSs ).
d. Now there you go: the twenty pure interlaced video segments above will be deinterlaced with SmoothDeinterlacer and then be encoded as Progressive ( progressive=true ).
Here's the thing. I'm looking through all my DVDs to try and find some actual video footage. So far I haven't been able to come up with any because even the TV shows I have on DVD are film based. I've observed that forced film material (ie had a hard telecince instead of 3:2 pulldown) can be converted to YUY2 as interlaced, but I'm not sure this is wrong since it was encoded interlaced to begin with. I have some more to post on this, but I have to collect some links and try to find the appropriate words to clarify colorspace conversions (mark it on your calender - I ran out of things to type!) so if anyone else has any comments on this, feel free to post them here.
|
|
Advertisement
|
 |
|
|
Staff Member
2 product reviews
|
6. August 2004 @ 15:28 |
Link to this message
|
Okay, I put together a little explanation of my points from the previous thread. If anyone finds any errors in this, feel free to post your corrections or PM me and I'll make any appropriate changes.
The difference between progessive and interlaced video is that in a Progressive picture, every line in a single frame is a representation of the same moment in time, while an interlaced frame represents 2 different moments approximately 1/60 of a second apart (1/50 for PAL). When you encode video to MPEG-2 it's essential to know what pixels are directly above and below each other. I'll use a diagram, in which each rectangle equals 1 pixel, to try to explain.
In the case of Progressive video, since there is only one picture, the encoder (or colorspace conversion utility) can assume that pixel 0,0 borders pixel 0,1 and 2,1 borders 2,3. For interlaced video this is no longer true. Since each interlaced frame is actually 2 pictures (fields) woven together, each with half the vertical resolution of the frame, pixel 0,0 no longer borders 0,1. In reality it now borders 0,2:
Let's say that in this small block of pixels we have 2 vertical lines that move exactly 2 pixels to the right from 1 frame to the next and then stop. In a Progressive source this would be 2 frames that look like this:
If the source is interlaced, the fields would be able to track the movement more gradually, since the period for each frame is half what it is for the entire frame, so the fields for the first frame could look like this:
While the second frame's fields might look like this:
As a result, the 2 frames, each being made up of 2 moments in time, would actually look like this:
Notice that the first frame, despite being made up of 2 straight lines, actually looks like a checkerboard pattern. Since there's no movement between the time the 2 fields in the second frame occur, they look just like the Progressive version.
The problem you have to worry about when encoding is that in order to get accurate results, the encoder needs to know what pixels border what other pixels. The errors you need to worry about don't occur based simply on whether the video is Progressive or interlaced, but rather whether the encoder knows they are Progressive or interlaced. When you set CCE to encode as interlaced what you're really telling it to do is treat the frame as 2 separate pictures that need to be encoded separately. As long as you do that you'll get an accurate picture. It's when you try to treat frames that are actually 2 interlaced fields (like the checkerboard that's really 2 straight lines) as Progressive that you run into problems. Since you're losing information (detail) by encoding to a smaller size you run the risk of having your straight lines changing shape to the point where you can't separate the fields and still have straight lines in each one.
The other issue is that if DVD-RB is in fact using Force Film settings on interlaced material (and as I mentioned I haven't done any interlaced encodes with it), the fact it's interlaced is the least of your problems. Unlike film, which has an actual framerate of 23.976fps and then duplicates fields to create interlaced video, in a video source all of the frames are unique. If you remove some of those fields you'll be missing information that's needed for smooth motion and you'll end up with choppy video. If I were to determine that it had done that on a video source I'd report it as a bug, not deinterlace.
|
|
erdoke
Suspended due to non-functional email address
|
8. August 2004 @ 04:31 |
Link to this message
|
I think this is a great explanation of a small issue of RB. Actually even I was able to get it. ;o)
In fact, I have never run into this in real life. (I backup my PAL videos with RB lately)
I agree that if RB works fine, there is no need to deinterlace at all. If not, it is time to report a bug to the author.
Today begins the rest of your life.
|
|
jdobbs
Senior Member
|
8. August 2004 @ 06:25 |
Link to this message
|
Quote:
The other issue is that if DVD-RB is in fact using Force Film settings on interlaced material (and as I mentioned I haven't done any interlaced encodes with it), the fact it's interlaced is the least of your problems. Unlike film, which has an actual framerate of 23.976fps and then duplicates fields to create interlaced video, in a video source all of the frames are unique.
As we discussed in another thread... DVD-RB doesn't use force-film. Force-film is an old method that distorts sources and can result in frame-loss and terrible jerkiness on true interlaced sources. DVD-RB keeps interlaced sources in their true form throughout the encoding process.
This message has been edited since posting. Last time this message was edited on 8. August 2004 @ 06:25
|
Staff Member
2 product reviews
|
8. August 2004 @ 06:44 |
Link to this message
|
|
Yeah, I posted this before we discussed it. I'm clear on it now ;-)
Edit: Actually the more I think about it, the more I like that solution. I'd have never thought of something like that, but it's so simple it's ridiculous.
This message has been edited since posting. Last time this message was edited on 8. August 2004 @ 06:50
|
|
Advertisement
|
|
|
|
jdobbs
Senior Member
|
8. August 2004 @ 06:59 |
Link to this message
|
|
I had a discussion on this subject with the Neuron2 and he has incorporated this capability into the newer versions of DVD2AVI and MPEG2ENC3 (the DGEncode versions).
This message has been edited since posting. Last time this message was edited on 8. August 2004 @ 06:59
|
|
