High Vs. Low Resolution Capture
Introduction
In general, there are two ways you can capture a video. You can capture in high resolution (XXX by 480 for NTSC or XXX by 576 for PAL/SECAM) or low resolution (XXX by 240 for NTSC or XXX by 288 for PAL/SECAM). Capturing in high resolution has a number of quality advantages, even if you are going to resize your video to a lower resolution. However, high resolution capture is much more taxing on your computer's resources. Below I will point out the advantages and disadvantages of choosing high over low resolution capture. I'll also talk about the importance of horizontal resolution during capture. Hopefully, this will enable you to make an informed decision when choosing a resolution for video capture.
Disadvantages of High Resolution Capture
1) Faster Hard Disk Requirements
A video that is 640 x 480 at 30 fps in YUY2 takes up about 18 MB per each second of video. It is possible to reduce that rate by using a lossless codec like Huffyuv, but the transfer rate is still relatively high. Because of this, I recommend using a newer 7200 RPM hard drive if you want to capture in 640 x 480. However, if you don't mind using a lossy codec (like PICVideo MJPEG), you can significantly reduce the required transfer rate. Unfortunately, this will also cause a drop in quality. Hardware compression capture cards will also lower the transfer rate but decrease the quality.
2) Larger Hard Disk Requirements
As I just mentioned, high resolution video capture requires a large transfer rate. This transfer rate corresponds directly to the amount of disk space you'll need to store a captured video. For example, if you capture a video in 640 x 480 at 30 fps in YUY2 with Huffyuv, the transfer rate might be 10 MB/sec. In this case, you would need 10 MB of free hard drive space for each second of video you wish to capture. You also need up to 176 KB/sec for the audio. When you take into account that processing a video can take up as much space as the original captured video, you'll see that many gigabytes of disk space can be necessary. Technically, it is possible to shrink your video during capture and processing through the use of software or hardware compression. However, the quality goes down as you increase the amount of lossy compression.
3) Higher CPU Requirements
You'll need a reasonably speedy processor if you wish to capture in high resolution. I recommend that you use a computer with at least a 500 MHz CPU. Although it is possible to capture in a high resolution with an even slower processor, you will probably have to make a tradeoff. For example, you can lower the CPU requirements by using a hardware compression capture card, but the quality will suffer. It's also possible to use a less demanding software compression codec (such as PICVideo MJPEG), but this will also decrease the quality. Finally, you can capture your video uncompressed to reduce CPU requirements, but you will need a huge, fast hard drive. For example, capturing uncompressed video at 640 x 480 at 30 fps in YUY2 requires a transfer rate of about 18 MB/sec.
Advantages of High Resolution Capture
If you think your computer is equipped well enough to handle high resolution video capture, here are the advantages:
1) Increased Detail
When you capture in high resolution, you are capturing more of the original signal. This obviously leads to an image with more detail. This is true even when capturing from a standard, low resolution VHS cassette. However, in this case the difference in detail isn't as great as there would be if you were transferring from a higher quality video source.
2) Reduction in Video Noise
This is one of the most important advantages to high resolution video capture. Analog video, by nature, contains video noise. Unfortunately, video noise is one of the worst things you can have when trying to encode a video. Noise increases the file size of an encoded video and decreases the quality. As I said above, when you capture in high resolution, you capture more of the original signal. This means that the ratio of intended signal to noise is higher when using high resolution capture. Therefore, the image has a reduction in noise. This fact is very useful when it comes to resizing a video from high resolution to low resolution. A video captured at 320 x 480 and resized to 320 x 240 will have less noise than a video captured directly at 320 x 240. (Note: 320 x 480 is not a typo -- see the next section about horizontal resolution for more information.)
3) Increased Smoothness in Interlaced Video
Many videos are in a format known as interlaced. In an interlaced
video, the odd and even horizontal lines in the image represent unique points in
time, as if they were separate frames. For this reason, interlaced video
can be thought of as running at 60 frames/sec (NTSC) or 50 frames/sec
(PAL/SECAM). When you capture in low resolution, most capture cards simply
cut out all of the odd or even lines. Therefore, low resolution capture
essentially removes half of the frames in your video, resulting in a jerkier
video when played on a television. High resolution capture, on the other hand, retains both the odd
and the even lines of the image. This keeps all the frames intact, making
your video play more smoothly.
Note: You will only see the increased smoothness of interlaced video when
watching it on an interlaced display, such as a television. If you view an
interlaced video on a computer monitor instead of a
television, you will see interlacing artifacts. For proper viewing of
interlaced video on a computer monitor, you will need to deinterlace your
video. (For more information about interlaced video, you can read my article
about Interlacing.)
4) More Accurate Inverse Telecine
Many NTSC videos are converted (telecined) from 24 frames per second (fps) to
30 fps so they can be watched on a television (which runs at 30 fps).
Using the inverse telecine process, it is possible to restore a video to its
original 24 fps state. This leads to a smaller, smoother video that
doesn't contains interlacing artifacts. However, the inverse telecine
process is difficult to accurately carry out. To remove the telecine, an
inverse telecine program has to analyze each frame of video for certain
characteristics. At times, it is difficult to find these
characteristics. When you capture in high resolution, the image has more
detail. This gives the inverse telecine program more information to work
with. Therefore, it can more accurately analyze the frames and remove the
telecine.
Note: For more information about telecined video, you can read my articles
about Telecining.
What About Horizontal Resolution?
Everything that I've mentioned so far relates to the vertical resolution of the video (the number of pixels from top to bottom). It is also important to discuss the horizontal resolution of a video (the number of pixels from left to right). First, let me answer a common question:
"Does capturing in a higher horizontal resolution (e.g. 640 x 240) and resizing to your desired resolution (e.g. 320 x 240) produce a video that has a higher quality than a video captured directly at your desired horizontal resolution (e.g. 320 x 240)?"
Usually, the answer is no. Unlike capturing in a higher vertical resolution, capturing in a higher horizontal resolution generally has little to no effect on video quality for most capture cards. Here's why:
Vertical Resolution of a Source Video
In an analog video signal, the image consists of a series of horizontal lines
that are stacked from top to bottom. The number of lines determines the
vertical resolution of the video. These lines are called scan lines, and
there is a defined number of them in every video. For NTSC video, the
image is made up of approximately 480 visible lines. For PAL/SECAM video,
the image is made up of approximately 576 visible lines. When you capture
a video in low resolution (XXX by 240 or 288), most capture cards simply cut out
every other line. This causes you to miss half of the detail in the source
image. By capturing in full resolution (XXX by 480 or 576), your capture
card retains all the lines of video. This gives you up to twice the detail
of low resolution capture.
Horizontal Resolution of a Source Video
In the horizontal direction (from left to right), an analog video signal is
not divided into a definite number of lines or pixels. Instead, the image
is purely analog. This means that it cannot be broken up into a certain number
of pixels that would contain all the detail in the original signal.
However, we can approximate the detail by taking numerous snapshots of
the source image at closely spaced intervals. This is exactly what your
capture cards does. It looks at the source video and takes snapshots,
called samples, of the image at closely spaced intervals. When it comes
time for the capture card to pass the video along to the rest of your computer,
it blends these snapshots together to create an image that meets your desired
horizontal resolution (e.g. 640 pixels). Essentially, the capture card is
capturing the video at a really high horizontal resolution and then resizing it
for you in hardware.
Most capture cards follow this same procedure no matter what horizontal resolution you specify. Therefore, an image captured at 640 horizontal pixels and an image captured at 320 horizontal pixels will both be captured internally by your capture card at a high resolution. The only difference lies in how much the capture card resizes the video. If you capture in a high horizontal resolution and then resize the video through software, you are usually wasting your time. Since the capture card will do the resizing for you, it is generally pointless to capture in a higher horizontal resolution than you want your final video to be. You won't gain any additional detail because the capture card internally captures the image at a high horizontal resolution, no matter what resolution you pick.
Of course, if your capture card doesn't do a good job of sampling or resizing the video, then you may get better quality by capturing in a higher horizontal resolution and resizing. However, I believe that capture cards like that are in the minority.