Interlacing

 

 

Introduction

Interlaced video is a common source of confusion for beginners in the world of digital video.  It requires a person to realize that how a video appears on television and how it is actually made up are quite different.  All video that is captured at full vertical resolution (480 or more horizontal lines) is susceptible to the effects of interlacing.  Thus, an understanding of interlacing is essential when dealing with full resolution video.  A familiarity with interlaced video is also necessary when dealing with telecined video.


Frames and Fields

In basic terms, a video can be thought of as being made up of numerous snapshots, called frames.  The frame rate, or the number of frames displayed each second, is 29.97 in the United States and other NTSC based countries.  For the sake of simplicity, we can round this number to 30 frames per second (fps).  In many European countries, they use a PAL or SECAM video system that displays exactly 25 fps.  For this article, I will base my explanations on 30 fps, but you can replace the number '30' with '25' for PAL/SECAM video and the same principles will hold true.

A television, however, does not deal with video in terms of frames.  Instead, it displays a video using half-frames, called fields.  Each frame contains exactly two fields.  One field is made up of the odd horizontal lines in a frame.  This is called the odd field or the top field since it contains the top line of the image.  The other field is made up of the even horizontal lines in a frame.  This is called the even field or bottom field.  Here is an illustration:

Fields in a frame

Since there are two fields in every frame, a television actually updates the display at 60 fields per second (or 50 fields per second for PAL/SECAM video).  Each field is displayed 1/60th of a second after the preceding field (or 1/50th of a second for PAL/SECAM video). 


Progressive Vs. Interlaced

Computer monitors and television screens do not display video in the same fashion.  A computer monitor updates the image only once for each frame of video.  Because of this, both fields that make up a video frame are shown simultaneously.  Thus, a computer monitor displays video at 30 fps.  This is called a progressive scan display.
A television, on the other hand, updates each field of a video frame separately.  Because of this, the television must update its display twice to show a complete frame.  It first updates the odd field, and 1/60th of a second later it updates the even field.  It continues to alternate between the odd and lines 60 times each second.  Thus, a television displays video at 60 fields per second.  This is called an interlaced display.

The field-based nature of video also allows a video to be recorded in one of two ways.  In the first method, called progressive video, the two fields that make up a frame represent the same instant in time.  This means that each entire frame also represents a single instant in time.  This is the preferred format for watching video on a computer screen because unlike a television, a computer monitor displays both fields of a frame at the same time.  
I've created a couple animations to help demonstrate how progressive video looks on a progressive scan display (monitor) and an interlaced display (television).  The top half of the ball represents a line of one field, and the bottom half represents a line of the other field.  In reality, the ball would be made up of many lines, but I've reduced it to two lines for simplification purposes:

Progressive video on a progressive displayTop = Field #1, Bottom = Field #2

Progressive video on a progressive display

Since both fields represent the same moment in time, the moving ball is complete in all frames.

Progressive video on an interlaced displayTop = Field #1, Bottom = Field #2

Progressive video on an interlaced display

Because an interlaced display only updates half of the image (one field) at a time, the moving ball gets split while field #2 catches up.


In the second type of recorded video, called interlaced video, each field represents a unique instance in time.  More specifically, each field of video represents a point in time 1/60th of a second later than the preceding field.  Because of this, a single frame of video does not correspond to a single instant in time.  Rather, half of a frame shows an object at one point in time and the other half of the frame shows the object 1/60th of a second later.
I've created some more animations to help demonstrate how interlaced video looks on an interlaced display (television) and a progressive scan display (monitor).  Once again, the top half of the ball represents a line of one field, and the bottom half represents a line of the other field.  In reality, the ball would be made up of many lines, but I've reduced it to two lines for simplification purposes:

Interlaced video on an interlaced displayTop = Field #1, Bottom = Field #2

Interlaced video on an interlaced display

Since each field represents a unique moment in time, the two halves of the moving ball never line up.

Interlaced video on a progressive displayTop = Field #1, Bottom = Field #2

Interlaced video on a progressive display

A progressive display must update both fields at the same time.  However, since each field represents a different moment in time, the moving ball is split across the fields.

 

So What Is Interlacing?

If you look closely at the four animations above, you'll see that only one of them doesn't split up the ball during motion (progressive video on a progressive display).  In the other three, the ball is split apart during motion.  However, when watching video on a television (interlaced display), the split isn't noticeable by human eyes.  This is because, among other things, a television doesn't produce as crisp an image as does a computer monitor.  Therefore, in only one of the four setups will the split be noticeable:  interlaced video on a progressive display (monitor).  This split in the image across fields due to motion is what is known as an interlacing artifact.

Now that we've identified when interlacing artifacts occur, let me show you an image that demonstrates interlacing in an actual video:

Interlaced image

If you look closely at the right side of the man's head, you will notice a series of horizontal lines.  These horizontal lines are an artifact of interlacing.  This is often referred to as the combing effect because the lines look like the teeth of a hair comb. 
The even lines in this picture, which are what you see on the right side of the man's head, make up the even field of the video.  The odd lines, which you can see on the left of his head, make up the odd field.  In this picture, the man is turning his head to the right.  The even field that you see on the right of his head represent a point in time 1/60th of a second after the time represented by the odd field.  The lines are visible because there was motion in the 1/60th of a second between the two fields.

To demonstrate how each field represents a different point in time, you can remove one field of video and blur together the lines of the remaining field.  This process, called deinterlacing, removes the interlacing artifacts.  Below you will find images of the above picture after I deinterlaced it by first removing the even lines and then the odd lines:

Deinterlaced by removing even lines

Image deinterlaced by removing the even field

Deinterlaced by removing odd lines

Image deinterlaced by removing the odd field

As you can clearly see, each field of video represents a different point in time.  In the first picture, the man is turning his head to the right and getting ready to open his mouth.  A fraction of a second later in the second image, the man's head has moved further to the right and his mouth has opened.


How Do I Remove Interlacing Artifacts?

If you're not going to watch an interlaced video on an interlaced display (like a television), you'll probably want to remove the interlacing artifacts.  There are three common methods that are used to eliminate interlacing artifacts found in interlaced video:

1)  Deinterlace the video.  There are three ways to do this:

A)  One way, as I demonstrated above, is to remove one field of a video (the odd or even lines) and then blend (interpolate) the remaining lines.  The downside to this method is that you lose half of the fields in your video.  This makes the video play less smoothly because the field that you remove represents a unique point in time.  For example, if I removed the even field in the above image, I would never see the image where his mouth is open. 
Additionally, deinterlacing reduces the resolution of a video because you are removing all of the odd or even lines.  Fortunately, you can use a VirtualDub Plug-in like "Deinterlace - area based" to partially work around this problem.  Instead of removing all of the odd or even lines, this plug-in only removes lines in portions of the video where interlacing artifacts are present.  In other words, it only deinterlaces areas where motion causes interlacing artifacts.  This is called adaptive deinterlacing.

B)  The second way to deinterlace video is to blend together the fields of the video so that combing effect of interlacing is not visible.  The advantage to this method is that since none of the fields are actually removed, you get a smoother playing video.  The disadvantage is that a "ghosting" effect will be present in your video during motion.  This is caused by the blending of the two unique fields within a single frame of video.  Here is what this type of deinterlacing looks like:

Image deinterlaced by blending the fields

As you can see, it looks similar to the source image except that the lines have been blended together to remove the interlacing artifacts.  The even lines on the right side of his head were blended with odd lines representing the wall behind him.  Because of this, the field with the open mouth looks transparent, or ghost-like.  The same is true with the odd lines on the left side of his head.
Using this method of deinterlacing can also cause a loss of detail due to the blending.  Once again, this problem can be partially avoided by using a VirtualDub Plug-in like "Deinterlace - area based".  This is an adaptive filter that will only blend the video in portions where motion and thus interlacing artifacts are present.

C)  The third way to deal with interlacing artifacts is through bob deinterlacing.  Bob deinterlacing is method by which each field of the video is made into a separate frame.  Because each field has only half the height of a full frame, each field is resized to twice its original height when it is transformed into its own frame.  Bob deinterlacing has the effect of removing interlacing artifacts while doubling the number of frames in the video.  Playing a bob deinterlaced video on a progressive display is similar to playing an interlaced video on an interlaced display because each field is updated separately.  However, there are a few disadvantages to bob deinterlacing:

i)  As I mentioned earlier, one field is made up of the even lines in a video and the other is made up of the odd lines.  Because of this, one field is offset vertically by a single line from the other field.  (For example, imagine that the top field contains lines 1, 3, and 5 and the bottom field contains lines 2, 4 and 6.  Then line 1 is offset from line 2 by one line, line 3 is offset from line 4 by one line, etc.)  To account for this one line difference, the top field pushed down by half a line and the bottom field is pushed up by half a line.  Unfortunately, this process isn't perfect, so some horizontal lines in your video may appear to flutter up and down.  To reduce this effect, you can use an adaptive bob deinterlacer like Smart Bob.  Adaptive bob deinterlacers preserve detail in areas with no motion.

ii)  You have to determine the field dominance of a video (described a few sections below) before you can properly bob deinterlace the video.

iii)  A bob deinterlaced video has twice the frame rate of the original video.  Playing back such a video will require more computer power.  Additionally, you'll need to re-interlace the video (returning it to the original frame rate) if you want to properly play it back on a television.

2)  Resize the video.  If you reduce the vertical resolution of your video by resizing it, you can eliminate the interlacing artifacts.  (You also have to reduce the horizontal resolution to maintain the correct aspect ratio.)  There are two manners in which you can resize a video to remove interlacing artifacts:

A)  You can simply remove one field (all of the odd or even lines) from a video and then squish the image horizontally.  This works because interlacing is only present in video that has two fields.  When you cut out one of the fields, the video frame now represents only a single instant in time.  Without the 1/60th delay between fields (since there is now only one), there can be no interlacing artifacts. 
This method of resizing is very similar to method "A" of deinterlacing.  The only difference is that instead of interpolating the remaining field to fill in the gap caused by removing one field, it simply produces an image that has half the vertical resolution of the source video.  After doing this, the resizing function also has to squish the image horizontally to maintain the correct aspect ratio.
The disadvantages to this method are the same as those for method "A" of deinterlacing.  You lose resolution and the video plays less smoothly due to the missing field.  There is nothing you can do to make up for this.

B)  Instead of removing one of the fields, you can blend the two fields together as you shrink the image to half its original horizontal and vertical resolution.  This is known as resampling the video to a lower resolution, and it is very similar to method "B" of deinterlacing.  The advantage is that since the fields are blended instead of removed, the video plays more smoothly.  The disadvantage is that ghosting artifacts will be present in areas of motion.  This is caused by the blending of the two fields and cannot be fixed.

3)  Capture in 1/2 vertical resolution.  This is the easiest way to prevent interlacing artifacts.  When you capture video at 1/2 vertical resolution (240 lines for NTSC, 288 lines PAL/SECAM), you are only capturing one field of video.  Since interlacing artifacts are the result of differences in the two fields in a single frame, there cannot be any interlacing artifacts if you have only one field.


When Should I Deinterlace?

As I mentioned earlier, each field in an interlaced video comes from a different point in time.  Therefore, interlaced video can be thought of as running at 60 fields/sec or 50 fields/sec rather than 30 fps or 25 fps.  An interlaced display is capable of displaying each field separately.  Because of this, an interlaced video will play much more smoothly than a progressive video when using an interlaced display.  Therefore, if you're planning to watch your output video on a TV (through an SVCD, for example), it's best to leave your video as interlaced.

On the other hand, if you want to watch your video on a progressive computer monitor, it's a good idea to deinterlace your video.  Because a progressive display updates both fields of a frame at once, you won't see the increased smoothness of interlaced video that occurs on an interlaced display.  Instead you'll see interlacing artifacts, which don't look very good.  That's why it's recommended to deinterlace your video if you'll be watching it on a computer.


Field Dominance

If you're going to play an interlaced video on an interlaced display, it's important to understand the concept of field dominance (aka field order).  As I discussed earlier, an interlaced display updates each field separately.  This brings up a question: which field should be updated on the display first?  Field dominance defines the answer to that question.

Because the fields represent series of progressing moments in time, there can only be one correct order in which to display them.  If the order is reversed, the video will quickly alternate between going forwards and backwards in time, leading to jerky motion.  The field dominance simply states whether the top/odd field or the bottom/even field is supposed to be shown first.  For example, a dominance of 'top field first' would mean that the top field is displayed before the bottom field in a frame.

In general, there is no consistent field dominance.  Each capture card and video source can lead to a different field dominance.  One way to determine the correct field dominance is to play a segment of your video on a TV.  If the motion is jerky, try changing the field dominance so the other field is displayed first (dominant).  You can also determine field dominance by using Avisynth.  See the Classifying Your High Resolution Video page for more information.  (Note:  If you're working with DV video, it's usually bottom (even) field dominant.)


Interlacing in Progressive Video

There are also times when a video that seems to be interlaced is actually progressive (non-interlaced).  There are two things that can cause this:

1)  The fields in your captured video are swapped or in the wrong order.
This is caused when your capture card puts the fields in the wrong order.  It can occur, for example, if your capture card starts capturing the odd/even lines in the wrong order or doesn't place them correctly into memory.
Fortunately, you can usually undo this type of interlacing and restore the original progressive frames using the internal VirtualDub filter called "field swap".  In some cases, you may need to use the Advanced Processing options of the VirtualDub Plug-in called "Smart Deinterlacer".  The process is completely non-destructive, which means that your video suffers no quality loss.

2)  The source video was telecined.  
Telecining
is a method by which progressive video that runs at 24 fps (such as a film) is converted to run at 30 fps (or 25 fps for PAL/SECAM video).
For NTSC telecining, also known as 3:2 pulldown, interlacing artifacts are caused by duplicate fields of video that are used to increase the number of frames displayed each second.  It is possible to remove these interlacing artifacts and restore the video to its original 24 fps with little or no quality loss.  The process can be a bit confusing, so I've written another article that deals specifically with Telecining.

In general, PAL/SECAM telecining does not use duplicated fields.  Instead, most 24 fps films are simply sped up by 4% to play at 25 fps on a PAL/SECAM system.  In some cases, during the conversion from 24 fps to 25 fps, one field of video is shifted by a frame.  This looks okay on an interlaced display (television), but it produces interlacing artifacts on a computer monitor.  You can rearrange the fields to their correct locations using the VirtualDub Plug-in called "Telecide" or the Advanced Processing options of the Smart Deinterlacer plug-in.  This process causes no loss of video quality.

Finally, in very rare cases a 24 fps video is converted to 25 fps for PAL/SECAM video by duplicating 2 fields over 24 frames to produce 25 frames.  Although I have never seen this method being used, I believe that it also causes interlacing artifacts.  You should be able to remove these artifacts by using the Telecide Plug-in for VirtualDub.