Technology Trade-offs
Since video is just a series of still images, and baseline JPEG encoders and decoders were readily available, people used baseline JPEG to compress real-time video (also called motion JPEG or MJPEG). However, this technique does not take advantage of the frame-to-frame redundancies to improve compression, as does MPEG. JPEG is symmetrical, meaning the cost of encoding and decoding is roughly the same. MPEG, on the other hand, was designed primarily for mastering a video once and playing it back many times on many platforms. To minimize the cost of MPEG hardware decoders, MPEG was designed to be asymmetrical, with the encoding process requiring about 100×the computing power of the decoding process.
Since MPEG is targeted for specific applications, the hardware usually supports only a few specific resolutions. Also, only one color space (YCbCr) is supported using 8-bit samples. MPEG is also optimized for a limited range of compression ratios.
If capturing video for editing, you can use either baseline JPEG or I-frame-only (intra-frame) MPEG to compress to disc in real-time. Using JPEG requires that the system be able to transfer data and access the hard disk at bit-rates of about 4 Mbps for SIF (Standard Input Format) resolution. Once the editing is done, the result can be converted into MPEG for maximum compression.
Quality Issues
At bit-rates of about 3–4 Mbps, “broadcast quality” is achievable with MPEG-1. However, sequences with complex spatial-temporal activity (such as sports) may require up to 5–6 Mbps due to the frame-based processing of MPEG-1. MPEG-2 allows similar “broadcast quality” at bit-rates of about 4–6 Mbps by supporting field-based processing.
Several factors affect the quality of MPEG-compressed video:
- the resolution of the original video source
- the bit-rate (channel bandwidth) allowed after compression
- motion estimator effectiveness
One limitation of the quality of the compressed video is determined by the resolution of the original video source. If the original resolution was too low, there will be a general lack of detail.
Motion estimator effectiveness determines motion artifacts, such as a reduction in video quality when movement starts or when the amount of movement is above a certain threshold. Poor motion estimation will contribute to a general degradation of video quality.
Most importantly, the higher the bit-rate (channel bandwidth), the more information that can be transmitted, allowing fewer motion artifacts to be present or a higher-resolution image to be displayed. Generally speaking, decreasing the bit-rate does not result in a graceful degradation of the decoded video quality. The video quality rapidly degrades, with the 8×8 blocks becoming clearly visible once the bit-rate drops below a given threshold.
Audio Overview
MPEG-1 uses a family of three audio coding schemes, called Layer I, Layer II, and Layer III, with increasing complexity and sound quality. The three layers are hierarchical: a Layer III decoder handles Layers I, II, and III; a Layer II decoder handles only Layers I and II; a Layer I decoder handles only Layer I. All layers support 16-bit audio using 16, 22.05, 24, 32, 44.1, or 48 kHz sampling rates.
For each layer, the bitstream format and the decoder are specified. The encoder is not specified, to allow for future improvements. All layers work with similar bit-rates:
| Layer I: | 32–448 kbps |
| Layer II: | 8–384 kbps |
| Layer III: | 8–320 kbps |
Two audio channels are supported with four modes of operation:
For normal stereo, one channel carries the left audio signal and one channel carries the right audio signal. For intensity stereo (supported by all layers), high frequencies (above 2 kHz) are combined. The stereo image is preserved but only the temporal envelope is transmitted. For ms stereo (supported by Layer III only), one channel carries the sum signal (L+R) and the other the difference (L−R) signal. In addition, pre-emphasis, copyright marks, and original/copy indication are supported.
FAQs
How do you determine which layer to use for highest sound quality?
To determine which layer should be used for a specific application, look at the available bit-rate, as each layer was designed to support certain bit-rates with a minimum degradation of sound quality.
Layer I, a simplified version of Layer 2, has a target bit-rate 192 kbps per channel or higher.
Layer II is identical to MUSICAM, and has a target bit-rate 128 kbps per channel. It was designed as a trade-off between sound quality and encoder complexity. It is most useful for bit-rates around 96–128 kbps per channel.
Layer III merges the best ideas of MUSICAM and ASPEC and has a target bit-rate of about 64 kbps per channel. The Layer III format specifies a set of advanced features that all address a single goal: to preserve as much sound quality as possible, even at relatively low bit-rates.
Insider Info
Layer III is also known as MP3, the popular music/audio standard.