Wavelet image coding with lattice vector quantization
Xianglin Wang, Guowei Wu, Lin Xinggang · 2002
Because of their attractive computational advantages over traditional LBG VQ, the two important lattice vector quantization schemes named PCPVQ and piecewise uniform LVQ are discussed and compared. However, in spite of their modified lattice distribution according to the source probability density function (PDF), the two methods are superior only over such a uniform lattice VQ coding scheme in which all lattices are grouped together with enumeration codes of the same length assigned to all lattices. A uniform LVQ image coding scheme named pyramid-shell uniform lattice VQ (PULVQ) is presented. Instead of grouping all lattices together, only lattices on the same pyramid shell are arranged as a group. As a result, different length enumeration codes are assigned to lattices of different groups so that more frequently occurring lattice vectors are assigned with shorter enumeration codes. Image coding results of both PCPVQ and PULVQ are presented. PULVQ outperforms PCPVQ because of its adaptive bit allocation.