Interlacing - explained in simple terms.

[Gew]

Hi!

I've been working with interlaced video for quite some while, but there are some fundamental stuff that has been bugging me for the last couple of nights.

So, I've read the Wikipedia article (http://en.wikipedia.org/wiki/Interlace) over and over, but there are still one fundamental question mark in my head. I suppose the answer is "implicitly implied" in eg. the Wikipedia article, but I've always had a rough time with grasping such things.

:stupid:

So, let me lay my question out simple. Take 576i50 video for example, my DV footage for such. So, every split second contains 25 frames, which further contains two fields (odd/even such). In total, 50 fields per second. This is de facto, no questions asked. But it's the actual capture / "shooting" within, that has gotten on my nerves.

When I record something with my DV cam, it first shoots a "half-frame" from it's even fields, and the very next instant it shoots another "half-frame" from the odd fields. Correct? This is my assumption, anyways.

So, I've processed simple math from this. 576/2 equals 288. So each field would contain 288 scan lines. Correct? This is my assumption, anyways.

Now comes the tricky part.

What I've failed to grasp, is how the actual capturing is done.

My first thinkable scenario is that the camcorder actually captures the first "picture" by every other field only. However, this sounds strange to me. This would mean that ie. one scan line is ,, well ,, actually omitted from each "visible picture" (field).

My second thinkable scenario is that the camcorder squeezes the whole capture area into the 288 scan lines worth of field x. Awaergh, not even sure how to explain it, even more unsure if someone who reads this understands my question in summary.

Awaergh. I bring an example to make my "mindflow" somewhat more reachable. Consider the following. I shoot some stuff. A 1px object is visible on scanline #123 in one instant, but only by the 1/50 of a second where scan line ...120,122,124... is being scanned. By the time ...119,121,123... is scanned (the next 1/50 second) this 1px object has moved through motion, eg. to scan line 124. Will this tiny little object be "lost in translation" (no pun intended), or is this where the magic of Persistence of vision (http://en.wikipedia.org/wiki/Persistence_of_vision) comes into play? I'm somewhat lost here.

I guess that I've become lost somewhere in the fundamentals, but I feel _this close_ to the final wrap of the whole concept. Just fill me in with the last piece of puzzle, and I'll be alright! :)

Too long, didn't read:
It feels strange that (if) interlacing would actually "drop" every other scan line of each field, i.e. picture detail would mos def be lost. Have I misunderstood it all? If one interlaced "picture" is made out of scan line 1,3 (etc), are these two somehow interpolated/blended to make suffer from the loss of scan line 2 less visible?

Oh, I really hope the answer I'll -- hopefully -- soon will be receiving will make everything clear, and not make me even more confused. Hehe.


Ty in adv~

[Guest]

Awaergh. I bring an example to make my "mindflow" somewhat more reachable. Consider the following. I shoot some stuff. A 1px object is visible on scanline #123 in one instant, but only by the 1/50 of a second where scan line ...120,122,124... is being scanned. By the time ...119,121,123... is scanned (the next 1/50 second) this 1px object has moved through motion, eg. to scan line 124. Will this tiny little object be "lost in translation" Yes, it will. Here's another example: You have a static horizontal line two pixels high. With interlaced sampling, it will appear to jump up and down.

It feels strange that (if) interlacing would actually "drop" every other scan line of each field, i.e. picture detail would mos def be lost. Spatial detail is lost compared to progressive scan. But consider this: if you do progressive scan at the same frame rate (not field rate) then you lose temporal resolution, because you have sampled half as many moments in time.

Interlaced sampling originated as a compression scheme. You could retain the temporal fluidity required for broadcasting sports (and other material requiring full temporal resolution) while transmitting only half the data.

[SomeJoe]

@Gew,

Yes, your analysis is exactly correct.

Another way to think about it is to forget the concept of "frames". Think of interlaced video as a repeating sequence of fields, where each field is an independent entity. Each field is independent in time from the two adjacent fields around it, and also is independent in space in that it captures spatial information not present in the two adjacent fields.

Indeed, an analog signal is itself only that -- a repeating sequence of F1 and F2 fields, without concept of a "frame".

It is only when the digitized fields are packed together into some sort of digital storage format that two adjacent fields are arbitrarily grouped together and stored together as a "frame". Yet it is improper to ever view this "frame" as a whole object, because the two fields that comprise it were not captured at the same time, but instead were 1/50th (PAL) or 1/60th (NTSC) of a second apart. That is also why we need a flag along with that "frame" that determines the field order, i.e. which field of this manufactured "frame" was captured prior to the other one. That information is needed for proper playback.

You're very correct that small (i.e. 2 pixels or less in height) objects will not have all parts of the object visible in both fields. 1 pixel high objects will only be present in one field. This is common in interlaced video, and the effect it produces on playback actually was given a name -- it's called line twitter. (And it has nothing to do with 140-character messages. :-) )

This is why it's always recommended to never use serif-style fonts for interlaced video, because the serifs on the letters are so thin that they always result in line twitter, making the text difficult to read.

Anyway, I'm rambling ... but this independent field concept (independent in both time and space) is a concept that many people, even video professionals, fail to grasp when dealing with interlaced video, and is probably the most important thing to keep in mind. It affects everything you do to interlaced video, from color subsampling to resizing to encoding.

[Gavino]

My first thinkable scenario is that the camcorder actually captures the first "picture" by every other field only. However, this sounds strange to me. This would mean that ie. one scan line is ,, well ,, actually omitted from each "visible picture" (field).
More or less, but as with all things in life (and especially with interlacing :)), the truth is a little more complicated.

I remember asking something similar when I was still a relative newbie.
At the time I discovered this post which I found useful. Also this article (http://www.adamwilt.com/TechDiffs/FieldsAndFrames.html).

[Gew]

Many thanks to all three of you, neuron2, SomeJoe & Gavino, for not "making a soup" out of it, but confirming in a simple manner. Really appreciated.

Now, the other night I laid there waiting for the sandman, and I happened to revisit the subject for interlacing. Suddenly it stroke me, that I had [in fact] grasped the pure nature quite some time ago.

What really(!) -- still -- puzzles me, and makes me confused, is the actual "drawing" slash "rendering" of interlaced video on interlaced displays. I've come to the conclusion that LCD television sets cant really handle interlacing, but instead approaches it with some sort of,, err,, what to call it, "artificial" slash "synthetic" variant, i.e. de(!)interlacing; whereas old school "fat tube" CRT television sets handles interlacing in a "natural way", due to it's nature. This, I (think) I know for a fact.

So, what kept me confused this night was the actual technical way a CRT tube draws the interlaced picture. My main concern was if the "first field" is still "hanging" when the "second" field is drawn, or if they "alternate alternately". I pictured different scenarios. Eventually, I decided I couldn't sleep, so I went up and started searching Google for some sort of easy-to-get illustrations of the process, which would hopefully bring clearness to this. I came across this excellent page, where the process is being demonstrated in a real "ADD friendly" / instructive way. Actually it was hosted on neuron2's own site.

http://neuron2.net/LVG/interlacing.html

Anyways, according to the illustrations on his site, the first field actually seems to be "hanging" meanwhile the second field is being drawn. This stroke me as odd though, since I find it sort of logical that this would introduce the same sort of combing as when watching interlaced video on a progressive display, i.e. two different moments in time drawn in the same frame. Do any of you guys follow my mindflow here? A wild guess is that the answer to this puzzle could lie somewhere in phosphor and "afterglow", but it's only wild guesses.

I'm actually lousy on explaining. So, I've actually managed to make my first GIF animations ever, hehe. Hope they explain the two "thinkable scenarios", and furthermore, that you guys could clear this out for me, and (if fig. 1 is correct) explain why it is that combing is not introduced on interlaced displays, even though there are actually an instant (2nd frame in the three-phase-illustration) where two fields [from different moments in time] are simultaneously present. So, here goes.

Figure #1
http://www.home.no/gewone/f1.gif
Figure #2
http://www.home.no/gewone/f2.gif

Ty in adv~
..and I'm thankfor for your patience with me! :)

[Guest]

The fields are scanned alternately by the electron gun. The phosphor has some persistence so the lines not drawn continue to glow a little bit. That means there is some combing visible on a real interlaced CRT. It's too small to be objectionable and most people never notice it.

[Gew]

:thanks:

I feel relieved.
Case closed!