Abstract
In a conditional replenishment coder for videotelephony only those picture elements (pels) which have changed significantly since the previous frame are transmitted. The amplitudes as well as the locations of these pels within the frame must be sent. In [2] locations are sent by addressing clusters of changed pels with respect to the beginning of the line. Increased addressing efficiency results if clusters of changes are positioned not with respect to the beginning of the line, but with respect to another cluster in the previous line or previous frame which has already been transmitted. Computer simulations using digitally stored videotelephone signals were carried out to evaluate such differential schemes. Results showed that differential addressing of this type and the use of variable word-length codes can yield about a 50 percent saving in bits required to address beginnings of clusters. The overall bit rate saving is much smaller, however, ranging from 10-15 percent depending on the efficiency with which pel amplitudes are coded.