eee508_VQ_part2

eee508_VQ_part2 - Vector quantization Example: Binary...

Info iconThis preview shows pages 1–5. Sign up to view the full content.

View Full Document Right Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon

Info iconThis preview has intentionally blurred sections. Sign up to view the full version.

View Full DocumentRight Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Vector quantization Example: Binary Splitting, M =8 centroid of the entire training data perturb centroid cluster w.r.t. new centroids compute new cluster centroids Advantage: Tree search can be used to reduce search complexity and computations EEE 508 Vector quantization Tree search VQ ( in general, does not have to be a binary tree) M =8 x 2 comparison vectors 4 i t 2 computations & 1 comparison 2 t t i & 1 i In general: r 1 r 2 r 3 r 4 r 5 r 6 r 7 r 8 4 comparison vectors 8 code vectors 2 computations & 1 comparison 2 computations & 1 comparison 6 computations & 3 comparisons 9 # computations = ( # splits at each node ) ( # stages ) 9 # comparisons = (# stages) For binary tree: # computations = 2 log 2 ( M ) = 6 for M = 8 1 t t Memory = where s = # splits at each node Advantage: Reduces search complexity and computations Disadvantages: Memory requirements => almost twice storage needed ; 1 1 stages # stages # 1-s s s s i i EEE 508 Disadvantages: Memory requirements => almost twice storage needed Vector quantization Exhaustive search VQ (conventional unconstrained) M = codebook size = 2 B N = Vector size N pixels/vector 1 2 N pixels Memory = M u N = 2 B u N : : M where B = bits per codebook index (no entropy coding) N B r vector per pixels # index codebook per bits # (sample) bits/pixel # rate bit EEE 508 Vector quantization N M N B r ) ( log t i l # index codebook per bits # (sample) bits/pixel # rate bit 2 Codebook size M = 2 rN exponential dependency on bit-rate r and vector size N N N vector per pixels # 1 N pixels Memory = 2 rN u N # distance computations = M = 2 rN # comparisons = M - 1 = 2 rN - 1 2 : : M EEE 508 Vector quantization...
View Full Document

This document was uploaded on 03/11/2012.

Page1 / 14

eee508_VQ_part2 - Vector quantization Example: Binary...

This preview shows document pages 1 - 5. Sign up to view the full document.

View Full Document Right Arrow Icon
Ask a homework question - tutors are online