Constrained storage vector quantization, (CSVQ), introduced by Chan anti Ge
rsho [2]-[4], allows for the stagewise design of balanced tree-structured r
esidual vector quantization codebooks with low encoding and storage complex
ities, On the other hand, it has been established in [9], [11], and [12] th
at variable-length tree-structured vector quantizer (VLTSVQ) yields better
coding performance than a balanced tree-structured vector quantizer and mag
even outperform a full-search vector quantizer due to the nonuniform distr
ibution of rate among the subsets of its input space, The variable-length c
onstrained storage tree-structured vector quantization (VLCS-TSVQ) algorith
m presented in this paper utilizes the codebook sharing by multiple vector
sources concept as in CSVQ to greedily grow an unbalanced tree structured r
esidual vector quantizer with constrained storage. It is demonstrated by si
mulations on test sets from various synthetic one-dimensional (I-D) sources
and real-world images that the performance of VLCS-TSVQ, whose codebook st
orage complexity varies linearly with rate, can come very close to the perf
ormance of greedy growth VLTSVQ of [11] and [12], The dramatically reduced
size of the overall codebook allows the transmission of the codevector prob
abilities as side information for source adaptive entropy coding.