| 摘要 |
A scalable quantizer is generated from a plurality of context models for each of a plurality of bit rates used in coding a source. A context model for a lower bit rate quantizes conditioning states when there is no compression gain in coding the image using child conditioning states in the higher bit rate context model over a parent conditioning state to which they may be quantized. The scalable quantizer includes a basic context model for a lowest anticipated bit rate and enhancement bits indicating how to adapt the basic context model to derive context models for higher bit rates. For image data transformed with embedded wavelet coding, context events are selected from corresponding or neighboring pixels at different resolutions and in different bit planes, and the order of the context events is optimized to reduce conditional entropy between the context events and the current value.
|
