# Singular Value Decomposition

Τεχνίτη Νοημοσύνη και Ρομποτική

5 Νοε 2013 (πριν από 4 χρόνια και 10 μήνες)

265 εμφανίσεις

Singular Value Decomposition

in Digital Signal Processing

By Tin Sheen

Signals

Flow of information

Measured quantity that varies with time (or
position)

Electrical signal received from a transducer
(microphone, thermometer, accelerometer,
antenna, etc.)

Electrical signal that controls a process

SVD background

The Singular Value Decomposition (SVD) of a rectangular matrix
A

is a decomposition of the form

A = U S V
T

U

and
V

are orthogonal matrices

and
S

is a diagonal matrix

The singular vectors form orthonormal bases, and the important
relation

A v
i

= s
i

u
i

shows that each right singular vectors is mapped onto the
corresponding left singular vector, and the "magnification factor" is
the corresponding singular value.

What is Digital Signal Processing?

Digital signal processing

(
DSP
) is the study
of signals in a digital representation and the
processing methods of these signals.

DSP includes subfields like: audio signal
processing, control engineering, digital image
processing and speech processing.

Noise reduction

The SVD has also applications in digital signal processing, e.g., as a
method for noise reduction. The central idea is to let a matrix
A

represent the noisy signal, compute the SVD, and then discard small
singular values of
A
. It can be shown that the small singular values
mainly represent the noise, and thus the rank
-
k

matrix
A
k

represents
a filtered signal with less noise.

SVD’s usefulness

SVD is sufficient and is the most optimal in
given image.

It is packed with energy in a given number of
transformation coefficients is maximized.

Easy to calculate.

Normal vs Noised Image

Block SVD filtering algorithm

Let the original, non
-
corrupted image F be represented as KxL matrix. Adding the noise to the
original F image will produces the noised image G of the same size

G = F + N

Where N = random KxL noise field

G will be divided into bxb matrix image

G
ij

= U
ij

S
ij

V
ij
T

where i = 1,2,3,4,…k; j =1,2,3,4,…,l

Where U
ij

is the left singular vector, S
ij

is the diagonal, and V
ij

is the right singular vector

From here we can calculate the mean value of the rank (rank obtained from G
ij

* S
ij
)

SVD’s calculation

The proposed algorithm has two basic steps: first, the noise variance
(
variance

of a random variable (or somewhat more precisely, of a
probability distribution) is a measure of its statistical dispersion, indicating
how its possible values are spread around the expected value) is
estimated,

And then filtering is performed on singular values and vectors. Noised
image is divided into square blocks of size
bxb
. For each block the singular
value decomposition is performed. In the consequent step, the average sum
of the last
t
singular values is calculated over all image

Noised calculation

Previously calculated SVD of image blocks will now be used for filtering.

First step in filtering is decreasing of noised singular value
s
i jr

for every image block :

ŝ
i jr

= s
i jr

p
1

* n
s

*w(r)

where ŝ
i jr

is filtered singular value,

p
1

is image dependent parameter and

w(r)
is a
weighting function that determines percentage of estimated noise variance to be
subtracted from noised

singular value

s
i jr

. The weighting function used in this work is chosen to be:

Implementation of DSP

Conclusion

Digital signal processing is a stealth technology. It is
the core enabling technology in everything from your
cell phone to the Mars Rover. It goes much further
than just enabling a one
-
time breakthrough product.
It provides ever
-
increasing capability; compare the