MEDICAL IMAGE PROCESSING
NOISE/ARTIFACT REMOVAL
CHAPTER 2
MEDICAL IMAGE PROCESSING
- Quality of medical images normally is affected and
degraded by noise
- Types of noise:
+ Random noises
+ Instrument noise
+ Physiological noises
- Causes for noise in medical images: various and broad
in level → high possibility to degrade the image quality
- Remove noise without losing information is a
challenge
-2-
2.1. Noise characteristicsMEDICAL IMAGE PROCESSING
( )
.p .d

−
= =
3
2.1.1. Random noises
- Noises by random processes like thermal noise in
electronic devices or photo counting
- Represented by
+ Pdf: probability distribution function pη(η).
+ Expectation: 1st moment
+ Variance: 2nd moment
- Medical image f(x,y) can be considered as a
representation of a random process f.
( ) ( )
( )
22
2.p .d
−

= − = −


2.1. Noise characteristicsMEDICAL IMAGE PROCESSING
-4-
- Gaussian noise
( ) ( )
2
x
x2
x
x
x
1
p x exp 2
2

−
=−

 

- Poisson noise
( ) ( )
k
P k exp . k!
= −
- Laplacian noise
( )
x
x2
x
x
2x
1
p x exp 2
2

−
=−



2.1. Noise characteristicsMEDICAL IMAGE PROCESSING
-5-
2.1.2. Structural noises
- Noise by X-ray grid
- Meshes or frames for image registration or alignment
- Labels to identify or localize patients, or to determine
image’s parameters
- Surgical implants like screws, splints
- 50/60 Hz noise: not popular in medical image