Multiple Description

Unlike layered coding (base+enhancement layers), multiple description coding creates independent descriptions. The more descriptions decoded, the larger the output quality.

In this simple examples, four balanced descriptions are created by separating odd and even rows, and taking every other pixel (a process known as polyphase downsampling). This is done in a pre-processing stage before encoding. The merge of decoded descriptions can also be done in a post-processing stage. Hence it is compatible with any standard, as MPEG-2.

In this example the number of packets and the aggregate bitrate is the same. When there are no losses the decoded quality favours single description. The opposite happens in case of losses. The resilience comes from the fact that it is unlikely to loose the same portion of the same image in all descriptions.

Fair comparison

It is not fair to compare single description, which is designed gor efficiency, against multiple descriptions, which is designed for resilience. A fair comparison is done with error-resilient single description so that the decoded quality in case there are no losses is the same.

In this example the single description is protected by reed-solomon (100, 90) forward error correction. Two descriptions are generated by separating odd and even rows.

The decoded quality when there are no losses still favours single description. However when the packet loss rate is above 5% multiple description takes over.

It must be noted also that, even when the average quality of multiple description is lower, it has also lower variance, which is very important for the user.

Foreseen applications

• Adaptation to varying bandwidth: the base station can simply drop descriptions/layers; more users can be easily served, and no transcoding process is needed.
• Adaptation to low resolution/memory/power: mobiles decode as many descriptions/layers as they can based on their display size, available memory, processor speed, and battery level.
• Easy picture-in-picture: with the classical solution, a second full-decoding is needed plus downsizing; with MDC/LC, it is sufficient to decode one description or the base layer and paste it on the display.
• Divide-et-impera approach for HDTV distribution: HDTV sequences can be split into SDTV descriptions; no custom high-bandwidth h/w is required.
• Enhanced carousel: instead of repeating the same data over and over again, different descriptions are transmitted one after another; the decoder can store and combine them to get an higher quality.
• Multi-standard support (simulcast without simulcast): descriptions can be encoded with different codecs (MPEG-2, H.263, H.264); there's no waste of capacity as descriptions carry different information.
• Easy cell hand-over: different descriptions can be streamed from different base stations exploiting multi-paths on a cell boundary.

Etc. 

Other schemes

An enhancement of the aforementioned scheme has been developed at University of Udine. Filters are used before downsampling.

As an example three descriptions can be created by separating odd and even rows, and taking also the average. Any two out of three description will suffice to decode the whole frame.

Another very clever scheme has been developd at Politechnic of Turin. It exploit one resilience tool of the new video codec H.264: redundant slices. Redundant slices are secondary slices based on the same reference as the primary slices. Typically the quantization step is increased, so that the secondary bitrate can be finely controlled.

Two descriptions are created by simply interleaving primary slices of odd/even macroblock rows and secondary slices of even/odd macroblock rows.