Watching interlaced video on a computer

Every once in a while a camcorder newbie turns up on a message board with a question: "Whenever there is a slight movement or panning the image breaks into series narrow horizontal lines. Is something wrong with the camera? Should I run back to a store and ask for replacement?" Those in the know respond with the usual: "Relax, these are interlace artifacts". Fine, but how to get rid of them? This article explains how to get rid of interlace artifacts when watching interlaced video.

What is interlaced video?

A television set creates image on the screen line by line. In traditional television all lines on the screen are not drawn or scanned in sequence from top to bottom, instead every other line is scanned. This method of display is called interlacing. With interlacing, an entire single image called frame is shown in two interlaced scans, each scan constitute a video field. Compared to non-interlaced or progressive television, interlacing allows using higher scanning rate within a given bandwidth. Higher scanning rate reduces flickering and allows to portray motion more naturally.

Interlaced scanning produces a visual defect called combing when objects in the image move around. With a traditional CRT-based television combing is not noticeable because fields are shown as series of fast flashes, caused by moving electronic beam, and also because scanning lines overlap a bit.

Modern LCD and plasma displays work differently than CRT displays and cannot display interlaced video natively. To avoid combing, interlaced video must be converted to progressive before it is presented to a viewer. The process of converting interlaced video into progressive is called deinterlacing.

All LCD and plasma consumer television sets have built-in deinterlacing circuitry to cope with interlaced broadcast signal. This means that if video is intended to be viewed on a TV, it does not have to be deinterlaced. It is a different story with computer displays. Computers have been working with progressive video for ages, and until recently they did not have built-in deinterlacing facilities, so interlaced video usually produces noticeable combing or ghosting when watched on a computer screen. Even at present time not all computers are able to display interlaced video correctly, everything depends on video hardware and software. Some video cards support hardware deinterlacing, but a proprietary video player or driver is usually required to make use of hardware deinterlacing.

In short, if video is intended to be viewed on a computer, it must be deinterlaced.

How to deinterlace

If you want to watch interlaced video on a computer you need to do one of the following:

• Use a video card with built-in deinterlacing. Most likely you will need a specialized video player that comes with the video card. For example, Nvidia advertises PureHD technology with all sorts of video processing that improve quality of video, but these features are available only if you use proprietary Nvidia video player.


• Use a video driver or codec that is capable of deinterlacing. Many codecs are capable of deinterlacing, but some are better than others. Many drivers can decode only specific video type, for example they may decode AVC but not MPEG-2. You may need to install separate drivers for different video types.


• Use a video player that has built-in deinterlacing engine.


• Convert interlaced video to progressive. This will allow you watching video on any computer with any hardware and uploading it to different online video sites.

Presently, converting interlaced video to progressive before distributing it is the best option to avoid interlaced artifacts.

Deinterlacing methods

Whether you decide to watch original interlaced video or to convert it into progressive, you need to know about basic deinterlacing methods. The most common are "bob", "blend", "single field" and "weave". Benefits and drawbacks of these methods are illustrated below using a frame from a native interlaced AVCHD video clip played through CoreAVC video codec, which provides manual deinterlacing controls.

Weave

"Weave" method simply joins -- or weaves -- two video fields together into one progressive frame. "Weave" will produce combing artifacts if used for native interlaced video. Notice that combing looks like "saw teeth" when video is watched at 100% size (image on the right). When video is downsized, combing may look like ghosting (image on the left).

This deinterlacing method is perfect if your video had been originally shot as progressive and recorded as interlaced by splitting one progressive frame into two fields. You may wonder why recording progressive video in interlaced format at the first place. This is common for 50Hz systems (PAL), where movies are first sped up 4% from 24 frames/s to 25 frames/s, then every frame is sliced in half and then resulting half-frames are broadcast as video fields. The same video equipment is used to broadcast theatrical releases as well as ordinary interlaced video. In television this technique is called 2-2 pulldown, in video this is called Progressive segmented frame.

The same technique is used in modern camcorders, for example the 30P mode of the Sony HC90/HC96, or the PF25 and PF30 modes of the Canon camcorders (PF24 is a totally different animal). When a "weave" method is applied to such a video, the resulting video preserves the frame rate (25 frames/s or 30 frames/s) and full vertical resolution of original video.

Do not use "Weave" for native interlaced video or you will get combing as shown above.

Bob

"Bob" should be used with native interlaced video. This deinterlacing method converts every field into a frame by doubling the lines, so sometimes it is aptly called "line doubler". Because every field is converted into a frame, the frame rate doubles from 25 frames/s to 50 frames/s (PAL) and from 30 frames/s to 60 frames/s (NTSC). This preserves the original refresh rate of interlaced video, keeping the fluidity of motion. This fluidity is what makes interlaced video looking like, well, video as opposed to "movie look".

With "bob" there is no ghosting, video looks sharp and clear at cost of loss of vertical resolution. This method is used by most consumer LCD and plasma television sets to handle interlaced video. Smarter TVs also apply "weave" method when they detect a static scene to achieve full vertical resolution.

On the flip side, "bob" deinterlacing may be too taxing for an average computer because of the doubled frame rate. Most video-sharing websites convert 50/60-fps video into 25/30-fps to decrease bandwidth. Presently, "bob" is not the best option for Web distribution.

Blend

This method may work well for videos with very little motion, but for high-motion video this is probably the worst deinterlacing method. Just like "weave", "blend" combines two fields into one frame, but to get rid of combing artifacts both frames are stretched vertically like in "bob" method. Resulting video looks soft, areas with motion exhibit noticeable ghosting, and the refresh rate is halved from 50 or 60 fields/s to 25 or 30 frames/s.

This method is not used in consumer television equipment in its raw form, more intelligent and costly methods are used instead. These more complex methods analyze motion between fields and blend fields accordingly, sometimes different parts of the fields are blended differently. Unless you have a really fast computer or a dedicated graphics card and appropriate software, you will not be able to use such motion-adaptive schemes. Instead, all that will be available to you is simple field doubling and blending.

The image above is a simulation, as CoreAVC codec does not provide "blend" deinterlacing option. But this is exactly the kind of image that you would see with, for example, InterVideo WinDVD media player. The InterVideo driver does not allow switching deinterlacing off or choosing another method, therefore interlaced videos watched with this player look soft and ghosty. To top it off, this player is not very fast and frame dropouts occur often.

Single Field

"Single Field" is the best deinterlacing method for video with fast motion that is targeted for Web distribution. With this method one field is thrown away completely, and another is converted to a frame like in "Bob" method, hence sometimes "Single Field" is called "Bob". In Sony Vegas editing application this method is called "interpolate". Adding to confusion, in older versions of CoreAVC this method was called "blend".

The major difference to "bob" is that refresh rate is dropped from 50/60 fields per second to 25/30 frames per second, so video looks less "live" and more "filmic". With "Single Field" both vertical spatial resolution and temporal resolution are halved, but the result is usually pleasing to the eye.

This method produces ghost-free video with Web-friendly frame rate. This is the best deinterlacing method for distributing over the Web.