Interlaced Field Rasters

Interlaced scanning was introduced in the 1930s as a way to reduce the number of lines that must be transmitted to half of what was required for progressive scanning. This kind of presentation requires that the odd- and even-numbered lines be presented independ-ently of one another. The presentation of interlaced lines is shown in Figure 5-2.

As the lines scan down the screen, the electron beam on a CRT display moves across the tube face, painting horizontal lines as it goes. When the beam reaches the bottom of the screen, it must go back to the top as quickly as possible to start the next downward pass.

This fly back accounts for about 8% of the scanning time, so you end up losing some of the lines.

Note how a half line is necessary on each pass in order to properly interleave the lines in the odd and even fields. It is tempting to just crop those two lines when com-pressing the video. Very often the choices you make with a video processing operation will have consequences (or knock-on effects) later in the workflow. In this case, removing just a single line has a knock-on effect that potentially compromises the field dominance. Lines must be removed in adjacent pairs. So those half lines are actually going to cost 4 lines of image data. An alternative is to keep them, and work out how to fill the other half of each line. A simple copy from an adjacent line would probably be the best solution.

So the picture area is traversed twice, with one field being painted first and then on the second pass, the other field is painted with the lines filling in the gaps. Figure 5-3 shows how an object is painted on the screen.

Odd field Even field

Visible lines

Horizontal flyback lines Vertical flyback Figure 5-2 Interlaced scanning pattern.

Whether the odd or even field is painted first depends on what is called the field domi-nance. It is very important that this is maintained identically to how the footage was filmed because if it is not, a strange “twittering” effect is observed and all the motion seems very jerky. That is because every 25th of a second, the time base goes back by a 50th of a second—a “two steps forward and one step back” scenario—and it is not pleasant to watch. Figure 5-4 plots the time base for a correctly clocked field presentation and one that has its field dominance reversed.

5.6.1 Persistence of Vision

The time base of the picture movement is in a 50th of a second increment, based on the field times and not the frame times. The field rate is twice the frame rate, but none of the lines in the odd field are displaying a part of the image that exists in the even field. It is as if there are two separate but simultaneous and totally synchronized movies playing as far as any motion is concerned.

Because the tube phosphor glows for a little while after the electron beam has passed over it and our eyes have a “persistence of vision,” you don’t see this as two alternate scans. Instead your eyes aggregate it into a single image.

Odd Even

Odd Even Odd Even Even

Odd Odd Interlace

Reconstruction

Figure 5-3 Interlaced field rasters.

5.6.2 Horizontal Fine Detail

Interlaced video demands that certain fine horizontal detail gets special treatment. If you draw a 1 pixel-wide line across the screen, it will appear to flicker because for half the time it is not being displayed. If you increase the thickness to 2 pixels, it appears to bounce up and down because it is present in alternate fields at different times. This is called twitter-ing. You must defocus the line in the vertical axis so it covers a slightly larger area, and the flickering effect is then reduced. See Figure 5-5.

The top line will flicker at 25 Hz because it only appears in one field. The second line will appear to bounce 25 times a second because it appears in a different place in each field. The third line will appear stable because either a full-intensity line is visible or two half-intensity lines are visible, but the average position is always the center line. Hence the eye is fooled into thinking it is stationary.

5.6.3 Object Motion Between Fields

Another downside to interlaced video is that motion carries on between fields. A ball bounc-ing past the camera will be in different positions on the odd and even fields. Figure 5-6 shows the distribution of the object as it moves.

If the original footage was shot on film and the telecine was transferred using a line-based algorithm rather than a pulldown, the effect is even grosser as shown in Figure 5-7.

Video time

Playback time

Orderly time line

Twittering time line Playback time

Video time

Figure 5-4 Time base twitter.

Pausing the video so that both fields are displayed presents a very disturbed image with some detail twittering about all over the place while the background remains still.

Some of these effects are alleviated by only displaying one field when the video is paused, but this sacrifices vertical resolution. Averaging between the two fields is another possibility that works some of the time, but it is a compromise as well. Chapter 33 revisits that topic in the discussion about de-interlacing. This is an important part of the practical encoding process.

1 pixel line

2 pixel line

Defocussed line

Position in first field

Position in second field shown in dashed outline

Odd field Even field Figure 5-5 Horizontal detail.

Figure 5-6 Interlaced object movement.