Databases Reference
In-Depth Information
7.
Use the test images Sena and Bookshelf1 to study the trade-offs involved in the selection
of block sizes in the forward adaptive quantization scheme described in Example 9.5.2.
Compare this with a more traditional forward adaptive scheme in which the variance
is estimated and transmitted. The variance information should be transmitted using a
uniform quantizer with differing number of bits.
8.
Generalize the Jayant quantizer to the nonuniform case. Assume that the input is from a
known distribution with unknown variance. Simulate the performance of this quantizer
over the same range of ratio of variances as we have done for the uniform case. Compare
your results to the fixed nonuniform quantizer and the adaptive uniform quantizer. To
get a start on your program, you may wish to use
misnuq.c
and
juquan.c
.
9.
Let's look at the rate distortion performance of the various quantizers.
(a)
Plot the rate-distortion function
R
(
D
)
for a Gaussian source with mean zero and
2
2.
(b)
Assuming fixed length codewords, compute the rate and distortion for 1-, 2-, and 3-
bit
pdf
-optimized nonuniform quantizers. Also, assume that
X
is a Gaussian random
variable with mean zero and
variance
σ
X
=
2
2. Plot these values on the same graph with
x
's.
(c)
For the 2- and 3-bit quantizers, compute the rate and distortion assuming that the
quantizer outputs are entropy coded. Plot these on the graph with
o
's.
σ
X
=
