Video Transmission for Third Generation Wireless Communication Systems

Journal of Research of the National Institute of Standards and Technology, March-April, 2001 by H. Gharavi, S. M. Alamouti

The basic block diagram of an interframe hybrid DCT video encoder is depicted in Fig. 6. According to this approach, a video frame is first divided into non-over-lapping blocks of 8 X 8 pixels, where each block is then DCT transformed, quantized (Q) and VLC coded. Except for the first video frame, which has to be intraframe coded (I-frame), the remaining frames may use a previously reconstructed frame known as the predicted or P-frame for motion prediction and compensation. At the cost of additional frame delays, both previous and future reconstructed frames may also be considered for motion prediction. This is known as bi-directional prediction, which has not been considered in our further elaborations.

For interframe prediction, a larger block of typically 16 X 16 pixels consisting of four neighboring 8 X 8 luminance DCT blocks--referred to as a macroblock (MB)--is used to perform block matching motion estimation and compensation. We note, furthermore, that a color MB also contains the so-called color difference components, which are processed at half the luminance resolution in vertical and horizontal directions. Since there are two color difference components, a MB can be viewed as though it was represented by six blocks. The estimated displacement motion vectors are multiplexed with the DCT coded data and transmitted as a part of the hierarchically ordered macroblock information.

The multiplexing structure of all existing video standards is generally based on a hierarchical, self-descriptive structure of the encoded parameters. For example, in the H.263 standard the video-coded information for each frame is arranged in four hierarchical layers. The top layer is the picture layer followed by a Group of Block (GOB) layer comprising a number of consecutive macroblocks, then the Macroblock layer, and finally a block layer. Each layer is furnished with some header information that may include synchronization bits such as picture start code, PSC, and GOB start code (for the two top layers), and that defines the nature of the information associated with each layer (e.g., inter/intra-type, quantization parameter, and motion vectors). If the header information of a specific video frame is lost during transmission, the decoder will have no indication as to how the frame, GOB, or MB has been coded. Therefore, any further data received will be undecodable, until the next PSC is recognized in the received bit pattern, e.g., by invoking correlation techniques.

As expected, the DCT coefficients associated with the particular video blocks are transmitted at the block layer and errors occurring in the DCT coefficients imply that the corresponding DCT coefficients are lost, since this information was variable length coded. If the transmission errors affected only higher frequency coefficients, the damage would be less catastrophic, since the more visually important low-frequency coefficients may have been recovered already. It is, however, important to protect the most error sensitive header information and as many lower frequency DCT coefficients as possible. The coding parameters hence have to be partitioned into a number of bit protection classes, in order to facilitate source-sensitivity matched error protection.