0% found this document useful (0 votes)
346 views5 pages

Max Lloyd Quantization

This document discusses quantization concepts for lossy source coding. It introduces L-level quantization and describes how to minimize errors in the quantization process. The Max-Lloyd quantizer partitions the dynamic range uniformly, but is not always the best solution. Considering the probability distribution of the source values can result in a minimum mean square error (MMSE) quantizer. The MMSE quantizer allocates more reconstruction values to more probable ranges in the source data. Iteratively solving for the reconstruction levels and decision thresholds can determine the optimal MMSE quantizer. As an example, a uniform distribution source is optimally quantized with a uniform quantizer. Quantization introduces distortion and is therefore a lossy step in source

Uploaded by

Rakesh Rt
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
346 views5 pages

Max Lloyd Quantization

This document discusses quantization concepts for lossy source coding. It introduces L-level quantization and describes how to minimize errors in the quantization process. The Max-Lloyd quantizer partitions the dynamic range uniformly, but is not always the best solution. Considering the probability distribution of the source values can result in a minimum mean square error (MMSE) quantizer. The MMSE quantizer allocates more reconstruction values to more probable ranges in the source data. Iteratively solving for the reconstruction levels and decision thresholds can determine the optimal MMSE quantizer. As an example, a uniform distribution source is optimally quantized with a uniform quantizer. Quantization introduces distortion and is therefore a lossy step in source

Uploaded by

Rakesh Rt
Copyright
© Attribution Non-Commercial (BY-NC)
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 5

Quantization Concept

UMCP ENEE631 Slides (created by M.Wu 2001/2004)

L-level Quantization

tmin

tmax

Minimize errors for this lossy process What L values to use? Map what range of continuous values to each of L values?
tmin tmax
(tmaxtmax)/2L

MaxMax-Lloyd Quantizer

Uniform partition
over a dynamic range of 2V Best solution?
X X

Maximum errors = ( tmax - tmin ) / 2L = 2V / 2L


tk tk+1

quantization error

Consider minimizing maximum absolute error (min-max) vs. MSE what if the value between [a, b] is more likely than other intervals?

Consider the Probability Distribution


UMCP ENEE631 Slides (created by M.Wu 2001)

Derivation of MMSE (Max(Max-Lloyd) Quantizer


z

p.d.f pu(x) t1 z z tL+1

MSE of L-level quantizer Optimal (MMSE) quantizer necessary conditions

r1

rL

Allocate more reconstruct. values in more probable ranges Minimize error in a probability sense
MMSE (minimum mean square error)
X X

assign high penalty to large error and to likely occurring values squared error gives convenience in math.: differential, etc.

An optimization problem
What {tk} and {rk } to use? Necessary conditions: by setting partial differentials to zero

MMSE Quantizer (Lloyd(Lloyd-Max)


UMCP ENEE631 Slides (created by M.Wu 2001/2004)

Observations on the MMSE Quanitizer

Reconstruction and decision levels need to satisfy

Solve iteratively
Choose initial values of {tk}(0) , compute {rk}(0) Compute new values {tk}(1), and {rk}(1)

For large number of quantization levels


Approx. constant pdf within t[tk, tk+1), i.e. p(t) = p(tk) for tk=(tk+tk+1)/2
Reference: S.P. Lloyd: Least Squares Quantization in PCM, IEEE Trans. Info. Theory, vol.IT-28, March 1982, pp.129-137

Example: Uniform Density


UMCP ENEE631 Slides (created by M.Wu 2001/2004)

MMSE Quantizer for Uniform Distribution


z

Uniform quantizer
Is optimal for uniformly distributed r.v. in MMSE sense MSE = 2 / 12 with = 2V / L
t1 tL+1

SNR of uniform quantizer


SNR = 6n dB
X

with

L = 2n

p.d.f. of uniform distribution

1/(2V) t1 2V tL+1

1 bit is worth 6 dB.

Uniform quantizer is the optimal mean square quantizer for uniform density.

z
UMCP ENEE739M Slides (created by M.Wu 2002)

Note: Quantization is a Lossy Step in Source Coding

Quantization is a lossy way to data compression


Lloyd-Max quantizer minimizes MSE distortion with a number of output bits n (i.e. a give L=2n)

You might also like