
|
 |
Video Production Hot Topics
The video topics are arranged in two categories. Each should be of interest to you when you are creating NetShow content, or for that matter, some of this information is helpful for any multimedia project. You can either select your specific topic of interest or just scroll through the list to see what's covered.
Video production tips
Third-party software
Whenever video, or any image for that matter, is reduced significantly in size, the detail, brightness, and contrast are reduced. A simple procedure that helps bring back detail, brightness, and contrast is to apply the Brightness and Contrast filter found in many video editing programs prior to the compression step. As a guideline, consider applying this filter to the video clip when the size is reduced by 1/3 or more. The following images show the effect of frame size reduction and then applying the Brightness (+20) and Contrast (-10) filter.
 Original 320 x 240 file |

Reduced to 160 x 112 |

Reduced to 160 x 112 and a Brightness (+20) and Contrast (-10) filter was applied prior to compression.
|
As discussed in the previous topic, reducing the frame size of any image causes a loss of detail. One way of increasing the image detail is to use the Sharpen filter in your video-editing program or through 3rd-party video filters. In Premiere, for example, you can apply a Sharpen filter prior to compression. However, this often results in images that look "too sharp" at the edges. A solution sometimes is to apply both a Sharpen filter and then a Blur filter at differing intensities. A one-step process is to try using a Gaussian sharpen filter.
The success or failure of compressing video data is dependent to a great degree on the amount of motion or scene change that is interpreted by the codec. When you apply transitions to a video clip, say a fade-in from black and then a fade-out to black, the codec interprets this as many steps of motion or scene change. This results in a less effective compression or at least having more key frames, which also increases the final data rate. The take home point is, do the transitions add to the value of the video? Don't use transitions indiscriminately. A second effect of using filters and transitions is that the time to compress the file is increased because more CPU-intensive processes are occurring when you build the file.
Video editing programs are just that; they edit video efficiently, and deal with audio only secondarily. Often you'll find that you can't perform the audio format compression or conversion that you need from within a video editing program; either for quality purposes or to reach a required data rate for your video. The solutions generally fall into two categories:
Export the audio track and use a dedicated audio editing program to do the fine-tuning whether that is file format conversion or compression. Then when you are ready to convert the .avi file into an .asf file using VidtoAsf, use the -audiofile filename.wav flag to replace the video track's original audio file.
Another solution to this dilemma is to use a program such as Sonic Foundry's Sound Forge 4.0b. This program, while primarily a powerful audio editing program, allows an .avi file to be opened directly, displaying the audio and video tracks. While the video track can't be edited, you can do far better audio file conversion and compression than from within most video editing programs. Yes, it's an extra step, but it might be the solution to your problem.
MPEG-4 requires the target data rate to be within a certain range when making a movie using certain frame size ranges. This chart summarizes the relationship between frame size and target data rate:
MPEG- Frame Size vs. Target Data Rate
|
Frame size |
|
Target Data Rate |
320 x 240 or larger |
> 176 x 144 or < 320 x 240 |
176 x 144 (QCIF) or less |
100 |
Y |
Y |
Y |
64 |
Y |
Y |
Y |
48 |
Y |
Y |
Y |
43 |
Y |
Y |
Y |
42 |
N |
Y |
Y |
32 |
N |
Y |
Y |
16 |
N |
Y |
Y |
15 |
N |
N |
Y |
1 |
N |
N |
Y |
The MPEG-4 configuration dialog box as displayed when working with on-demand content contains three controls: key frame interval, smoothness/crispness slider, and data rate slider. Optimization depends on the type of video and the target data rate. The type of video is determined by a "semi-objective" evaluation of the amount of motion and scene change in the video; commonly the type is categorized as low-motion, medium-motion, or high-motion video. The target data rate is the limiting factor when it comes to getting video that works for your needs.
Using the NetShow Encoder provides more options for configuring the MPEG-4 video codec, or for that matter, other video codecs that you might be using. The additional configuration options found in the NetShow Encoder are Pixel Format and Delay Buffer size.A comprehensive discussion on the configuration settings for video codecs is available in the document, Microsoft NetShow ASF Encoder Video Quality Options.
Remember, these settings are starting points; the type of video you're working with and what's acceptable and unacceptable video quality are personal opinions. But the first thing you should do is set the target data rate; the following chart gives an idea of how the target data rate setting, the type of video, the frame size, and the actual data rate are related. Each data rate value is linked to the actual .asf file, so you can look at each video and make your own evaluations. If you do not have Microsoft Media Player installed, go to the Microsoft NetShow site and download Media Player.
Codec comparison chart:
The next step, and the easiest one, is the key frame interval. Use the default value as a starting point. This value is optimized for each particular codec. Here are some typical settings:
Codec
|
Key frame interval (default) |
MPEG-4 |
1 per 3600 seconds; set in codec configure dialog. If the key-frame interval is set to 0, all frames are key frames. |
Indeo Raw |
No compression; can't set key-frame interval |
Indeo YVU9C |
Can't set key-frame interval |
VDO |
1 per 9999 frames |
Indeo 3.2 |
1 per 4 frames |
Indeo Interactive 4.1 |
1 per 15 frames |
Cinepak |
1 per 15 frames |
Lastly, the smoothness/crispness value needs to be set. An understanding of what this value is attempting to control helps you select a value. Smoothness refers to the smoothness of the motion; it is related to frame rate presentation. Crispness refers to the detail or sharpness of the video image. These two values are exclusive, meaning that if the value is increased for one quality, for example crispness, the quality of the other value, smoothness of motion, is decreased. You can't have your motion and image quality too!
Take a look at these video examples using different smoothness/crispness values.
MPEG-4; Smoothness (0%)
MPEG-4; Smoothness/crispness (50%)
MPEG-4; Crispness (100%)
Tough question; what one person calls low-motion, medium-motion, and high-motion video and how the codec interprets this motion might be quite different. There probably isn't one "magic bullet" codec for all types of video at multiple data rates. But here are some guidelines:
MPEG-4 works best across a range of video motion from low-motion, low-bit-rate video to high-motion, intranet-bit-rate video. It is clearly the codec of choice for working with a variety of video motion types; however, for video that contains a constant and definable amount of motion, other choices might be considered. MPEG-4 is designed such that the images are crisper and moereblocky appearing than with VDO, for example. This works excellent for low-motion type video. However, when the motion rate increases, this blockiness may be less visually appealing.
H.263 is a low-bit-rate video codec optimized for low-motion, talking head-type video. For this type of video, the image and motion quality may be the best of the three video codecs mentioned here. However, if the video contains a sudden combination of low and high motion video scenes, this is not the codec of choice. While the low-motion scenes might be high quality, the higher-motion video quickly becomes unacceptably smeared or pixelated.
VDO, on the other hand, is designed to compress the video in such a way that the images appear smoother and softer. This gives less definition for low-motion video such as talking heads. However, with high-motion this smoothness often works better than the crisp, blocky appearance associated with MPEG-4.
One other characteristic of video besides motion and scene changes is the visual texture; for example, a video with a snow skier on a slope, with a heavy background of trees and a partly cloudy sky, has a large amount of texture. The smoothness of the snow and the details of the trees between the snow and sky make a complex, contrasting texture which various codecs handle differently. MPEG-4 is quite good with this type of complex texture video, while VDO makes the video look very soft, losing much of the detail. Another issue is the final data rate; MPEG-4 is optimized for high-texture scenes while VDO is optimized for low-texture scenes. This results in a lower final data rate for MPEG-4 compressed video than with VDO.
In summary, for low to high-motion video or video with high-texture scenes, MPEG-4 is an excellent starting point; VDO is certainly an option, however. If the video exhibits medium to high-motion or is primarily low texture scenes, VDO is probably a better starting point than MPEG-4. For low-motion, talking head-type video, H.263 is an excellent choice. The one point to keep in mind is how consistent is the motion factor in your videos; if it is variable, MPEG-4 is probably your best choice. However, your preference might be different, so I suggest that you do some tests with different types of video you commonly work with and determine what is visually most appealing to you.
A comprehensive study on various computers and CPU utilization is available which provides many details related to encoding machine selection, different types of video encoded, codec selection and configuration, and so on. See CPU Utilization Study.
That's easy -- you don't! VidtoASF converts the .avi or .mov (QuickTime) file to an .asf file with a few options such as changing the audio file used, setting the amount of buffering time at the beginning of the file, and so on. But the data rate is determined at the video compression stage when the .avi or .mov file is being made. If you have to change the data rate of the video file, you must return to the videoediting program that produced the video clip and recompress the file at another target data rate.
In on-demand created video content, the ability to scan forward and backward through the video clip by moving the slider on the Media player is determined by whether enough key frames are present and which codec is used to compress the video. VDO, Duck, Cinepak, Indeo 3.2, and Indeo 4.1 all allow scanning if you use the codec's default key-frame interval. On the other hand, MPEG-4 doesn't allow you to seek when you use the default key-frame setting. By decreasing the key frame value you can have the video to seek.
There is one way, however, of allowing a video to be "seekable" without altering the key-frame values. When the .avi or .mov file is converted to an .asf file using VidtoASF, set the seekable flag to on. For example:
Vidtoasf -in filename.avi -leadtime 5000 -seekable on
Content produced with NetShow Real-Time Encoder as part of NetShow v2.0 normally doesn't create seekable .asf files. ASFChop creates an index on .asf files longer than 10 seconds so that you can seek (similar to fast-forwarding) through the file. To do this, specificy the -in and -out commands without any time, and be sure to give the output .asf file a different name than the input .asf file.
Content produced with NetShow Encoder as part of NetShow v3.0 integrates the ASFChop indexing functionality. Indexing of .asf files can be done automatically from within NetShow Encoder. Since this is potentially a CPU-intensive process, this option is in the Local File options property page or in the Custom Configuration wizard. The default setting is on and the procedure is started upon completion of encoding the .asf file.
Yes. A dialup line with a good connection should be adequate, though using 33.6-Kbps modems at each end would be better and using ISDN would be optimal. Keep in mind, however, that just because you have a modem capable of providing a 28.8-Kbps connection, this doesn’t mean that you will always get all the potential bandwidth. Due to network traffic, network "noise," or other factors, it is not uncommon to see no more than 24-26 Kbps connections on a 28.8-Kbps dial-up connection.
This is one of the main reasons that the suggested total bandwidth for ASF content targeted for a 28.8-Kbps connection should be no more than 23-24 Kbps.
Third-party software:
Premiere 4.2 currently does not allow you to make a movie using the MPEG Layer-3 or Vivo audio codec. The solution is to select another audio codec, make the movie, and then use one of three approaches:
- Separate the audio and video tracks using a program such as AVIEdit or export the audio as a .wav file from within Premiere. Edit the audio in an audio editing program using the MPEG Layer-3 codec. Now, using VidtoASF, build the .asf file with both audio and video tracks using the -audiofile flag. This takes the video file and the audio file and re-assembles them into one .asf file. One important point to remember: if audio synchronization is important, do not change the audio file's content, as that doesn't give the same audio synchronization as the original video. Here is the most basic example:
VidtoASF -in filename.avi -audiofile -filename.wav
- Use Sonic Foundry's Sonic Forge 4.0b or later to open the .avi file. This allows you to refine your audio codec choice and retain the previous video compression work. You can, however, change both the audio and video compression schemes from within Sound Forge.
The Microsoft NetShow Tools includes the NetShow Plug-in for Adobe Premiere. This allows you to easily export AVI files from Premiere to the ASF format.
MPEG-4 requires the frame size to be evenly divisible by 16. For example, a frame size of 160 x 120 gives this error, while a frame size of 160 x 112 or 160 x 128 build without incident. The solution for this problem is easy; change your target frame size to be evenly divisible by 16.
Third-party applications may handle how they display codecs differently. For example, one video editing application might display every codec installed on your machine whether it can actually use it or not. Another video editing application may display codecs more intelligently; whether the codec is displayed is dependent on whether the codec can actually be used on your system and whether all your output settings are appropriate for the codec.
An example of how two applications handle the display of codecs differently is described below.
A Windows 95 system has the following video codecs installed as displayed in the Advanced dialog box of the Multimedia Control Panel applet.
Cinepak
Indeo video interactive 4.1 by Intel
Indeo video 3.2 by Intel
Indeo video RAW YVU9 by Intel
Microsoft H.263 video Codec
MPEG-4 Video High Speed Compressor
VDOnet VDOWave Video Codec
Adobe Premiere recognizes all codecs except #5, Microsoft H.263 video codec. Regardless of the video settings, this codec does not appear in the codec selection list. In Corel Lumiere, however, if a correct frame size is used, the codec is displayed. For example, if you use a frame size of 160 x 120, the codec does not appear, while if you use 160 x 112, the codec does appear.
|
|