Progressive Transmission of Motion Sequences at Extremely Low Bit Rates

As an example of application of bit allocation analysis for progressive transmission of moving targets at very low bit rates, we propose the Rational Embedded wavelet VIdeo Coding (REVIC). It follows three key features (see Table II):

• Attention-based quantizer formation which compacts points of maximum attention to a small number of quantizers, and which generates only one quantizer isolating coefficients to points with low attention value, [7]. Then, instead of performing temporal filtering along the motion trajectory, the proposed coder simply uses more restricted languages to represent moving patterns: It compacts peaks of the points of maximum attention to a small number of high-attention quantizers by attention-value thresholding points of maximum attention.
• Intra-quantizer prioritization which follows the embedded zerotree coding scheme and provides substantial coding gains over the first-order entropy for significance maps.
• Optimized inter-quantizer prioritization which follows bit allocation analysis. Since the present truncation time $t$ of the progressive transmission might be the last one, bit allocation analysis does not permit accumulation of idle bits which should be transmitted in future times. The complete transmission of a motion sequence involves sequential considerations, but this may essentially reduce to repeated analyses based on the static model at each specific time instant. A user of the system may, at different bit rates, choose the strategy for computing profit vector $p$ in a way that is most appropriate to the application of interest.

Table II:

Key Features

Attention-based quantizer formation instead of motion filtering

Intra-quantizer prioritization by embedded zerotree coding

Inter-quantizer prioritization by bit allocation analysis