H.263

The bandwidth requirements of H.261 are considered high for the majority of domestic and mobile users, so a new standard was developed for bitrates near 30 kbps. The H.263 encoder is fundamentally the same with the H.261 encoder, as shown in Figure 2.6. However, some modifications were incorporated in the recommendation to improve the video quality at low bit rates, including:

  

Figure 2.6: A simplified block diagram of an H.263 encoder

 
• The use of new input video formats, as sub-QCIF ( $128 \times 96$), 4CIF ( $704 \times 576$) and 16CIF ( $1408 \times 1152$).^2.2
• The use of half-pel prediction in block matching, unlike H.261, where candidate blocks are displaced only to integer pel positions, and motion estimation is coarser.
• Unrestricted search range for motion vectors, allowing better prediction near the frame boundaries.
• Advanced prediction mode, which assigns different motion vectors to each of the four $8 \times 8$ blocks comprising a macroblock.
• Arithmetic coding rather than Huffman encoding.
• PB-frames, which are combinations of a forward and a bidirectionally predicted frame, see the next paragraph for definition of P and B frames.

These features make the codec much more complex than the respective H.261 implementations, but, with the continually improving hardware available at continually decreasing prices, they can be justifiable, especially for mobile users and terminals.

Hosted by www.Geocities.ws