Scribd
Upload
11 views25 pages

SC 13

This document discusses non-uniform quantization, where the quantization intervals are made smaller in regions of high probability density and larger in regions of lower probability density. This allows the input to be approximated better in high probability regions, at the cost of worse approximations in low probability regions. An example non-uniform quantizer is shown with smaller intervals near the origin for sources like Gaussian, Laplacian or Gamma distributions. The reconstruction point in each interval is the centroid of the probability mass in that interval.

Uploaded by

Warrior Bro
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
11 views25 pages

SC 13

This document discusses non-uniform quantization, where the quantization intervals are made smaller in regions of high probability density and larger in regions of lower probability density. This allows the input to be approximated better in high probability regions, at the cost of worse approximations in low probability regions. An example non-uniform quantizer is shown with smaller intervals near the origin for sources like Gaussian, Laplacian or Gamma distributions. The reconstruction point in each interval is the centroid of the probability mass in that interval.

Uploaded by

Warrior Bro
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPT, PDF, TXT or read online on Scribd
You are on page 1/ 25

Non Uniform Quantization

• If the input distribution has more mass near the origin,


the input is more likely to fall in the inner levels of the
quantizer.
• Recall that in lossless compression, in order to minimize
the average number of bits per input symbol, we assign
shorter code words to symbols that occurred with higher
probability and longer code words to symbols that
occurred with lower probability.
• In an analogous fashion, in order to decrease the average
distortion, we can try to approximate the input better in
regions of high probability, at the cost of worse
approximations in regions of lower probability.
• We can do this by making the quantization intervals
smaller in those regions that have more probability mass.
• If the source distribution is like Gaussian, Laplacian,
Gamma etc. we would have smaller intervals near the
origin.

• If we want to keep the number of intervals constant, this


would mean we would have larger intervals away from
the origin.

• A quantizer that has nonuniform intervals is called a


nonuniform quantizer.

• An example of a nonuniform quantizer is shown in next


slide.
x1 x2 x3 x4 x5 x6 x7

x1
1, , …. -1

-1
• The reconstruction point for each
quantization interval is the centroid of the
probability mass in that interval.

• The decision boundary is the midpoint of


the two neighbouring reconstruction levels.
~
• Next slides : Instead of yk s, x (sk ) are used.
• Instead of fx(x) ---- p(x) is used.
Problem

• A continuous-valued random variable X has a probability


density function (pdf) given by:

X is the input to a two-level (1 bit) non- uniform scalar


quantizer which has initial decision thresholds
t0 =0, t1 =0.5 and t2 = 1.0.
Design the quantizer such that SNR is greater than 10 dB.

You might also like