Abstract: A data compression process and apparatus using adaptive coding of a digital signal by vector quantization being applicable to transmission of digital audio signals and digital television signals as well as other signals. Quality of the received signal under the constraint of limited transmission bit rate is improved by selective optimized local replenishment of the code-book. Selected high distortion cells are identified using the current code-book and the partitioning of the selected cells reconfigured interdependently with reconfiguration of the neighboring cells to replenish the code-book, and replenishment data transmitted to the receiver. The preferred vector quantization includes classification of the cells according to distortion criteria, the code-book being a concatenation of subcode-books which are separately replenished.