Redefining the video world 1: Interlacing versus Progressive

Posted on June 22, 2011 at 15:30 by Kevin Chung in Technology

In a previous post I’ve made an announcement about presenting a new concept/method to preserve the visual quality of video. In this new series of blogposts I’ll be highlighting some of my ideas. In all of these topics I will express the importance of keeping things simple.

“Simplicity is brilliance.”

The series will consist of the following topics, which will be posted over the next weeks (one post per week):

• Part 1: Interlacing versus Progressive
• Part 2: Format war
• Part 3: Telecine
• Part 4: NTSC versus PAL
• Part 5: Television manufacturers

At times it all may sound I’m expressing myself from the perspective of a purist. Please note that purists can see both flaws and opportunities. They can be a driving source of power to reach new levels of innovations ready for you to be used!

Today we are still using technologies from the past that are outdated and do not fit the way we currently work with videos. It spreads like a disease through all stages of video processing. We need to get rid of those outdated technologies or upgrade them to fit within the current standards. When this has been taken care of new opportunities for innovation will present themselves and the road is clear!

Beware that the discussion here is only a representation of my vision as an encoding engineer that puts effort into creating a rich end user experience!

PART 1: INTERLACING VS PROGRESSIVE
In quite a few of my previous blogposts I’ve expressed my views on interlacing. The reason I start with this topic is because I believe this technology is the Achilles’ heel in video world. The invention of interlacing was necessary in the past because TVs were displaying pictures using alternate lines or say fields. Interlacing was also a nice way to keep temporal resolution for better motion fidelity at the cost of half the spatial resolution at limited bandwidth.

Today interlacing doesn’t add anything beneficial to the way we process videos, be it production, mastering or watching videos. Cameras at their core capture images progressively. Current TVs show pictures in progressive mode, be it LCD or plasma technology. Yet today we are still using interlacing or derivatives thereof. Look at TV broadcasting and commercially released DVDs. But even Blu-ray didn’t get away with this. The days of never having the need to deinterlace seemed like a utopia to me.

Interlacing should be blown into oblivion
I’m ready to believe that the fact that interlacing is still used nowadays is in fact a political issue. Engineers would agree with me that interlacing is useless, but we are not in control of this market here. Interlacing should be blown into oblivion.

It would be a great start for the upcoming H.265/HEVC standard to not support interlacing anymore. If this happens I will consider that a bigger revolution than going from Standard Definition to High Definition.

To summarize why interlacing is a sin today.

1. Interlacing throws away half the spatial resolution and interpolates the missing fields to yield a progressive frame during playback. This is data and computationally inefficient and hurts spatial quality. Modern codecs perform much better on progressive video. The addition of interlacing alone makes the total code length unnecessary long. Take X264′s recent development for example. Revision R1956 – R1991 alone was entirely dedicated to make MBAFF interlacing possible in X264. As a free encoder it was already very good in feature set. With the addition, X264 is even more complete now. However X264 developers at their core don’t want anything to do with interlacing for the obvious reasons.
2. Interlacing is not fool proof. Uneducated engineers or consumers easily make fundamental mistakes when processing interlaced video. Common mistakes are encoding interlaced video as progressive video and thus not deinterlacing beforehand, or resizing interlaced video, which is even worse.
3. Interlaced video needs to be lowpassed in order to avoid typical side effects like flickering, twitter, etc. This softens the picture in most cases and therefore you lose details. These side effects are typical for interlacing and don’t occur with progressive video.
4. Interlaced video must be deinterlaced in order to view it on a progressive scan display. Deinterlacing, however, is never perfect and becomes computationally expensive when resolution increases. It’s not linear to resolution, more exponentially. The equipment we use today to view videos are progressive scan. Deinterlacing is a horrible process that’s mandatory if we want to watch videos today. As I said earlier: Interlacing is easy. Deinterlacing is a pain.
5. For post processing your videos it is often better for filters and effects to be applied on progressive video. If not you need to deinterlace and then re-interlace (if needed). Inefficient.
6. Interlacing is the foundation of all kinds of unnecessary difficult methods like telecine, 2:3 or 2:2 pulldown and “cheap” NTSC/PAL standards conversions and vice versa. Bad reversal methods before editing ultimately destroy the video to an irreversible state after going through several video editing companies.

How does this reflect in a daily end user scenario? In most cases seeing artifacts because of the limitations of interlacing is simply a blow to the experience of video. Nothing is more annoying than seeing horizontal borders having artifacts. That could be caused by resizing of an interlaced video, which are very susceptible for artifacts. Or how about subtitled animation that has aliasing? Most certainly this resulted from bad treatment of interlaced footage.

3D vs interlacing
There’s more but these few outlined points alone cover pretty much the whole spectrum. To come to an end for this blog post, I want to use a recent development as an example to illustrate the evilness of interlacing that affects the whole video chain or anything related to it.

3D is hot. Television manufacturers are really pushing this into your home. There’re several methods for creating the 3D illusion to our perception. But it is clear that for a good 3D experience you need progressive pictures. Interlacing has no business here. And to my bitterness interlacing is still involved in some way with 3D!

LG’s previous method was just to use one of the 2 alternating fields for the left and right eye, e.g. odd fields for the left eye and even fields for the right eye. Now they want to give both eyes all the fields. Both are fundamentally wrong. No compensation method is going to hide or fix this. Get the fundamentals right first!

For the 1st method each field represents a capture at a different time domain. For 3D you want pictures from a different angle at the same time domain. They also don’t state if the fields are deinterlaced…?

For the 2nd method they use all the fields for both eyes. This is a problem too as it involves some mixing of frames across two different time domains for each eye. This is complete nonsense. It’s fundamentally not correct.

Television manufacturers should really put their R&D focus in other areas. But that’s beyond the scope of this blog post. In part 5 you can read more.

The entire video world is a much better innovative place if we get rid of interlacing. Progressive is fundamentally the basis for capturing images. It’s simple, it’s foolproof, and it works!

And as said: simplicity is brilliance!