why do we use deinterlacing filters?

[deadrats]

i know this sounds like a truly stupid question, but why do we use deinterlacing filters? consider this for a second, if we have an progressive scan source video and we want an interlaced output, we don't use an interlacing filter, we simply set the encoder to encode as interlaced and viola, we're done.

keeping that in mind, the progression during the transcoding process is source --> uncompressed intermediate --> target, in the process of decoding the source, aren't the frames automatically structured as full frames (not fields).

as an experiment i loaded an interlaced vob into tmpg express 4, disabled deinterlacing and set the mpeg-2 encoder to progressive, the result was a file were i can't see any tearing or interlacing artifacts.

is there a flaw in my logic somewhere? if i'm wrong shouldn't there be some evidence in the outputted file?

[poisondeathray]

i know this sounds like a truly stupid question, but why do we use deinterlacing filters? consider this for a second, if we have an progressive scan source video and we want an interlaced output, we don't use an interlacing filter, we simply set the encoder to encode as interlaced and viola, we're done.

keeping that in mind, the progression during the transcoding process is source --> uncompressed intermediate --> target, in the process of decoding the source, aren't the frames automatically structured as full frames (not fields).

as an experiment i loaded an interlaced vob into tmpg express 4, disabled deinterlacing and set the mpeg-2 encoder to progressive, the result was a file were i can't see any tearing or interlacing artifacts.

is there a flaw in my logic somewhere? if i'm wrong shouldn't there be some evidence in the outputted file?

The flaw in your logic is you're encoding as interlaced, but the content isn't necessarily interlaced. Encoding as progressive or interlaced doesn't necessarily have any bearing on the actual content. You can have interlaced content encoded as progressive, or progressive content encoded as interlaced. There are even other bizarre forms like field shifting, hybrid interlacing, various pulldown cadences. Also your playback software might be processing it (e.g. it might be deinterlacing automatically, so it's probably not the best way of testing)

One definition of interlacing is each field is a separate moment in time. If you have a 60p sample (frames per second), then you can make a 60i sample (fields per second). Otherwise you don't have that temporal resolution. If you start from a 30p sample, your program is probably duplicating frames (each field doesn't represent a new moment in time, so the content isn't interlaced)

Also, don't confuse re-interlacing with deinterlacing. The first is generating interlaced content from a progressive sequence. The latter is generating progressive content from interlaced content. Notice in both cases the content is important, not necessarily how it was encoded

[edDV]

It starts with the source and the desired output. What comes in between is work flow.

If your source is interlace (e.g. typical DV, HDV, AVCHD camcorder) and desired output is DVD or Blu-Ray, there is no reason to deinterlace.

If the source is progressive, there is no need for deinterlace.

If the source is interlace and you need progressive output (e.g. web stream), then you need to deinterlace at some point in the process.

In a typical case source is interlace and output needs to be various formats. In that case, edit an interlace master then convert that to various formats.

[jagabo]

Because sometimes video needs to be deinterlaced. Like when uploading to Youtube.

Your VOB file may have been encoded interlaced but it did not contain interlaced frames. This is very common with PAL DVDs.

Attached is a real interlaced, BFF, mpeg 2 elementary stream file (M2V) for you to play with.

[Bonie81]

all the experts have already been replied in details.
choice between interlaced or de-interlaced depends on on your target container and targeted viewers.
if viewers are gonna watch over hardware player it's no need to De-interlace.
if viewers are gonna watch on youTube or a computer monitor screen, it is better to de-interlace.
visit this link, you will notice major difference on first pic and very last pic on your PC monitor.
but, on hardware player with nice video bitrate probably first one will look more crispy.
bff.m2v posted bt jagabo has been de-interlaced, so u can see the difference.
Frankly i can't see much difference on my PC screen.

[manono]

i know this sounds like a truly stupid question, but why do we use deinterlacing filters? consider this for a second, if we have an progressive scan source video and we want an interlaced output, we don't use an interlacing filter, we simply set the encoder to encode as interlaced and viola, we're done.

You're in the US so I doubt you'll be encoding your progressive sources as interlaced (unless progressive 29.97fps, possibly). If for DVD, film is 24fps. You encode that as progressive 23.976fps (not interlaced). But DVD requires interlaced 29.97fps output. How are you going to get to 29.97fps? You apply pulldown. The encoder is not set for interlaced encoding. Perhaps 99% of the film sourced DVDs are done this way.

[deadrats]

Your VOB file may have been encoded interlaced but it did not contain interlaced frames. This is very common with PAL DVDs.

this is all very confusing, i thought i understood what was going on but it's giving me a headache (well that and the half bottle of wild turkey i buried).

how is it possible to have a file that's encoded as interlaced but doesn't actually contain interlaced frames? my understanding of interlaced vs progressive is thus:

with a progressive frame, all the pixels are independent of one another and they are all displayed on the screen at one time, simultaneously. for instance 720x480p means that you have 720 pixels across and 480 pixels high that are all displayed at the same time.

with an interlaced frame, the pixels are arranged in lines, with each line being a separate entity. the lines are displayed in alternating fashion, 1 at a time, until the full frame is displayed. for instance a 720x480i consists of 480 lines and each line has 720 pixels and depending on whether it has a top field first or bottom field first bias, the frame is displayed as line 1,3,5...and then 2,4,6... for tff and the other way for bottom field first.

is the above not correct? if it is correct how then is then possible to encode a frame as interlaced but not actually be interlaced? do you mean that a flag is set in the video stream that tells the media player to display a progressive frame as interlaced? if that's the case isn't possible to do the same with interlaced frames, set a flag to tell the player to display it as progressive?

i downloaded the sample you provided and for the life of me i can't see the interlacing, i played it back with gom player and kmplayer and at first i thought that i must have "deinterlacing" enabled within the player, checked, nope disabled, i checked in the nvidia control panel to see if i have deinterlacing enabled globally, nope, disabled, yet i don't see the sample as interlaced. media info and gspot say it interlaced but when played back on a computer monitor (which are progressive displays), shouldn't interlacing artifacts be apparent when deinterlacing is turned off?

[poisondeathray]

kmplayer deinterlaces mpeg2 video automatically

You can check individual FIELDS in avisynth (by using separatefields() ) , or open in vdub and you should see combing. When you open a video in vdub, you're looking a frame, not a field.

Interlaced CONTENT means each consecutive FIELD is a different moment in time . As you step through individual FIELDS, each one will be different. A,B,C,D,E,F

Progressive CONTENT will have same pair FIELDS A,A,B,B,C,C (i.e. each consecutive pair represents the same moment in time)

CONTENT and ENCODING can be different. You can flag a stream as interlaced, and you can even change the encoding algorithm as interlaced (e.g. motion vectors are different for fields vs progressive frames)

*You still have to look at individual fields to identify with certainty what the sample is. Just opening it up in a non deinterlacing player or vdub isn't enough. For example, if you have a phase shifted sample , it will show combing, but the content is progressive.

[jagabo]

A frame of video is just a 2d array of pixels. A field of video is every other scan line of the frame. The top field comprises all the even numbered scanlines, the bottom field the odd numbered scanlines. So every frame of video has two fields -- whether it's progressive or interlaced. With a progressive frame both fields come from the same picture, the same point in time. With an interlaced frame the two fields come from two different pictures, two different points in time. When you watch an interlaced video on an interlaced TV you see one field at a time. By the time one field is being drawn the other has faded away. You never see an entire frame on an interlaced TV.

A 30 fps progressive video is stored as 30 frames every second and intended to be viewed as 30 different pictures every second. A 30 fps interlaced video is stored as 30 frames per second but is intended to be viewed as 60 fields every second.

Logically, when a video is encoded progressive the frame is compressed as a whole. When a video is encoded interlaced the two fields are separated and compressed individually. Upon decoding the two individual fields are woven back into a frame and handed to the editor or player.

Computers, of course, aren't really designed to work with interlaced video. So to properly view interlaced video on a computer you need to deinterlace. Either before encoding or during playback.

[budwzr]

"Interlace" is an old Hollywood hack that has to do with matching voltage frequency in old style TV set electron tubes to avoid flicker.

It has no other purpose.

[edDV]

"Interlace" is an old Hollywood hack that has to do with matching voltage frequency in old style TV set electron tubes to avoid flicker.

It has no other purpose.

For a given resolution and motion sample rate, interlace permits transmission at about half the bandwidth vs progressive hence its use for broadcast. That is why 1080i is still used for TV transmission.

[budwzr]

Isn't it because they need to match the frequency of the tube TV's that have analog (composite) converter boxes?

[edDV]

Isn't it because they need to match the frequency of the tube TV's that have analog (composite) converter boxes?

All "NTSC" derived video is transmitted at a 59.94 update rate. Interlace sends 59.94 fields per second (480i or 1080i), progressive is sent at 59.94 frames per second (480p or 720p). ATSC also has provision for 480p/720p/1080p at 23.976 frames per second but nobody is using it currently because older TV sets can't receive it without a set top box. The 59.94 rate is maintained for backwards equipment compatibility.

Internally, cable and dbs companies are beginning to use 23.976 fps for up/down sat links for some movie networks but they aren't yet offering that as a set top box output. Instead you get 59.94 with either interlace telecine or progressive 3-2 frame repeats.

In PAL markets all video is sent 50 fields per second interlace or 50 frames per second for 720p again for backwards compatibility. Most PAL movies and TV series are native progressive (i.e. shot on 24p film) but are sped up to 25 fps and reinterlaced for transmission at 576i or 1080i.