Everything you ever wanted to know about video processing, but figured the answers would bore you to tears.

Scaling, upscaling, upconverting, deinterlacing, jaggies, 3:2 pulldown. Chances are you’ve never been able to figure out at least one of these words. No problem. Everyone has to learn somewhere. I’ll admit it: I was not, in fact, born with this knowledge. I know there are those that claim they were, but I suspect some of them are lying.

What I aim to do with this guide is to distill down the heady jargon and explain what all these words mean and why you should care (or not).

For demonstration purposes, let me introduce Norman, the beagle.

Why use a cartoon dog to demonstrate video processing? Well, to be honest, because he cracks me up every time I look at him.

To be even, ah, honester, test patterns are about as exciting as herding cats. Sure, there may be a few moments of interest, but ultimately, everyone gets bored.

No matter how well you describe these terms, though, it’s infinitely easier to see what's being described with images.

So Norman will be our guinea pig, so to speak. And in our first experiment, we are going to quite literally tear him in half.

Deinterlacing

There are two types of video: interlaced and progressive. Interlaced means that half the image is shown every sixtieth of a second. Progressive means the whole image is shown every sixtieth of a second. All modern TVs are progressive. Well, almost all. If you still own a CRT or one of a few odd Hitachi plasmas, you can skip this section as your TVs are interlaced and don’t need to “de” anything.

Basic cable channels and most HD channels are interlaced. If you look at Norman in Figure 1, you can see him interlaced:

Figure 1: Interlaced video (480i and 1080i) splits each video "frame” into two “fields.” The first field consists of the odd lines in the image, while the second field is the even lines. Your TV combines these two fields back into one frame: 240+240=480 or 540+540=1080. The lines are exaggerated for ease of understanding and mild humorous effect.

The first field is the odd lines of the image. With SD, this is 240 lines. With HD, it’s 540. Then one-sixtieth of a second later, the even lines show up. Your TV then combines these and shows them to you as one. To do this, it needs to deinterlace the image.

I guess you could also say “make progressive.” Same number of syllables.

If the image isn’t moving, this is easy. Like sliding two combs together (something I haven’t owned in well over a decade), the lines from the first field and the lines from the second field combine to create a full frame of video. Easy.

What’s not easy is if the object on screen is moving. In Figure 2, Norman is sliding very quickly across the screen:

Figure 2: If the subject has moved between the moments when the camera captures the image (as Norman has above), the resulting image, without processing, can look a little... off as the TV combines two fields that are not similar.

When the video camera captured his image for field 1, he was in one place, but when it captured him one-sixtieth of a second later for field 2, he had moved. So now, when your TV goes to combine the two fields, Norman looks a little Picasso.

Not good.

Good video processors know this is going to happen. So they take a look at the two fields, see that something has changed, and attempt to smooth out this difference. How and how well they do this varies by company.

If we refer to “tearing” or “combing,” we are describing an artifact we see in the image that is caused by poor deinterlacing.

3:2 Pulldown

We get an extra level of difficulty when the original content is from a movie. Nearly all films are shot at 24 frames per second. Your TV runs at 60 frames per second (some new TVs are 120 or more; we’ll get to them later). In order to show content that is 24 frames per second on a device that is 60 frames per second, some fancy shuffling has to be done.

As you can see, what we’ve done with Norman in Figure 3 is what’s called a 3:2 pulldown.

Figure 3: The same is true with content that has been converted from film. The 3:2 (or more accurately, 2:3) sequence doubles and then triples each film frame. If a TV doesn’t detect this sequence and tries to combine the fields, the frames created from the third “B” frame and the first “C” frame will look like Norman from Figure 2. If you click on the image, you can see a larger version.

The first frame of film is split into two fields, the second frame of film is split into three fields, the third frame of film is split into two fields and so on. Yes, it would be more accurate to call it 2:3 pulldown, but old names die hard. Just ask Bruce Willis.

So when your TV recombines these fields, it creates a frame out of the first two fields (no problem), another frame out of the second two fields (still good), and then a frame out of one field from the second film frame, and one field from the third film frame (crap…).

What the TV needs to be able to do is detect that there is this 3:2 sequence and compensate, essentially throwing out the third field. What you’re getting on screen then is two video frames of the first film frame, then three video frames of the second film frame and so on to 60. This is fairly common with current-model TVs, though even just a few years ago, getting anything to pick up the 3:2 sequence with HD material was exceedingly rare.

Some TVs aren’t 60 frames per second (otherwise known as 60 hertz [Hz], meaning cycles per second). Some displays operate at 72, 96, or 120 Hz—all multiples of 24—which makes the conversion a little easier. These displays can (though not all 120 Hz models do) ditch the 3:2 sequence and just show each film frame more than once.

In the case of 120 Hz models, this is five times. Keep in mind that while this will remove some of the jerkiness, or “jutter” in the image when you have a 3:2 sequence because the original content was film, it will never be as smooth as something shot on video. Personally, I like this “film look.” Others don’t, which is why you can find all 120 Hz TVs with some sort of frame interpolation. This technology actually creates entirely new frames to stick in between the original film frames, smoothing out the motion.

If a TV doesn’t pick up the 3:2 sequence, you may get the “tearing” and “combing” we mentioned before. In addition, you may get a moiré pattern, or worst case, the TV will give up, and will actually just show you half the vertical resolution, doubled to fit the screen. No current models I know of do this, but some used to.

Now it’s time to stretch Norman out.