Developments with a Zerotree Audio Codec
Ben Leslie, Chris Dunn, Mark B. Sandler · 1999
INTRODUCTION scales. It uses this property to significantly reduce the number of zero bits which would be transmitted if pure The concept of embedded-zerotree quantisation was bit-plane encoding were used. The EZ quantisation first introduced by Shapiro [1] as Embedded Zerotree process is composed of two stages which are repeated Wavelet (EZW) coding for image compression. This successively until the target bit-rate has been met: was shown to provide compression performance equal to or better than alternative techniques, with very low a. Significance: compares the coefficient algorithmic complexity. In addition, the process matrix with a threshold level, and produces a fully embedded code, where bits in the produces a sequence of bits or symbols bitstream are produced in order of importance. The which identifies which coefficients are coding process may thus be terminated at any point, significant with respect to (ie greater than) yielding a reconstruction quality that is roughly the threshold. The threshold is halved proportional to the length of bitstream produced. An following each pass. alternative way to consider an embedded code at a b. Refinement: produces a series of bits certain rate is that it includes all lower-rate codes. This which halves the uncertainty interval of property is very useful for loss recovery and scalability, coefficients which have already been asshownin[1]. foundtobesignificant.