Competence Centre on Information Extraction
and Image Understanding for Earth Observation
2011 IEEE International Geoscience and Remote Sensing Symposium
26/07/2011
Vancouver, Canada
SATELLITE IMAGE ARTIFACTS
DETECTION BASED ON COMPLEXITY
DISTORTION THEORY
Avid ROMAN GONZALEZ
Mihai DATCU
Avid
Roman
Gonzalez
Prof.
Dr. Mihai
Datcu
Competence Centre on Information Extraction
and Image Understanding for Earth Observation
OUTLINE






The Artifacts, problematic.
Rate-Distortion Function
Kolmogorov Complexity
Kolmogorov’s Structure Function
Experiments and Results
Conclusions
Avid
Roman
Gonzalez
Prof.
Dr. Mihai
Datcu
Competence Centre on Information Extraction
and Image Understanding for Earth Observation
ARTIFACTS :
 The artifacts are artificial structures
that
represent
a
structured
perturbation of the signal. Therefore,
these artifacts induce errors in the
indexation of the images.
Avid
Roman
Gonzalez
Prof.
Dr. Mihai
Datcu
3/47
Competence Centre on Information Extraction
and Image Understanding for Earth Observation
Aliasing
Saturation
Avid
Roman
Gonzalez
Prof.
Dr. Mihai
Datcu
Strips
...
Blocking
4/47
Competence Centre on Information Extraction
and Image Understanding for Earth Observation
Data cleaning, or data cleansing or scrubbing
- Detecting and removing errors and inconsistencies
from data in order to improve the quality of data .
Data quality problems are present in single data
collections, such as files and databases, e.g.:
- Due to misspellings during data entry.
- Missing information or other invalid data.
Avid
Roman
Gonzalez
Prof.
Dr. Mihai
Datcu
5/47
Competence Centre on Information Extraction
and Image Understanding for Earth Observation
 Is to predict or determine the existence of defects,
to model it, and then design a method to detect and
correct them. For example we have the lines
correction methods presented by [Hyung-Sup Jung
2009].
 Specific Artifacts.

Using data compression techniques to implement a
method more generic PARAMETER FREE regardless
the type or model of artifact.
Hyung-Sup Jung, Joong-Sun Won, Myung-Ho Kang, and Yong-Woong Lee, “Detection and
Restoration of Defective Lines in the SPOT 4 SWIR Band”, Transaction on Image
Prof.
Dr. Mihai Datcu
Processing,
2009.
Competence Centre on Information Extraction
and Image Understanding for Earth Observation
 The Rate-Distortion (RD) Function is given by the
minimum value of mutual information between
source and receiver under some distortion
restrictions.
R( D)  min QQD I ( p, Q)
 The RD function shows how much compression
(lossy compression) can be used without loss of
distortion preset value.
Avid
Roman
Gonzalez
Prof.
Dr. Mihai
Datcu
7/47
Image
JPEG
Lossy
Compression
Image cf 1
Image cf 2
Image cf 2
.
.
.
Image cf n
Decompression
Competence Centre on Information Extraction
and Image Understanding for Earth Observation
Images with different
compression factor (cf)
+
-
Features
Vector
(compression
errors)
Classification
 For the artifacts detection, we propose to use the RD
function obtained by compression of the image with
different compression factors and examine how an
artifact can have a high degree of regularity or
irregularity for compression.
Avid
Roman
Gonzalez
Prof.
Dr. Mihai
Datcu
8/47
Competence Centre on Information Extraction
and Image Understanding for Earth Observation
Kolmogorov Complexity
K x  is the length of a
shortest program to
compute x on a universal
Turing machine
K(x) is a non
calculable
function
K  x   min q
qQx
String 1
000000000000000
15 x (Write 0)
String 2
1010001010111011
Write 1010001010111011
Avid
Roman
Gonzalez
Prof.
Dr. Mihai
Datcu
9/47
Competence Centre on Information Extraction
and Image Understanding for Earth Observation
Kolmogorov’s Structure Function
 An approximation of the RD curve using the Kolmogorov
complexity theory could be the Kolmogorov Structure
Function (KSF).
 The relation between the individual data and its
explanation (model) is expressed by Kolmogorov’s
structure function.
 The original Kolmogorov structure function for a data x is
defined by:
hx    min log S : S  x, K (S )   
S
 Where:
S is a contemplated model for x.
α is a non-negative integer value bounding the
complexity of the contemplated S.
Avid
Roman
Gonzalez
Prof.
Dr. Mihai
Datcu
10/47
Competence Centre on Information Extraction
and Image Understanding for Earth Observation
Kolmogorov’s
Discrimination
Structure
Function
for
Texture
 To evaluate the behavior of the KSF for different textures,
we use de Brodatz images databases. We show the
textures used for this experiments.
Avid
Roman
Gonzalez
Prof.
Dr. Mihai
Datcu
Competence Centre on Information Extraction
and Image Understanding for Earth Observation
5
10
x 10
9
8
7
6
5
4
3
2
1
0
0
1
2
3
4
5
6
5
x 10
 We can observe that the KSF can discriminate more or less
the different structure, the curve KSF has a similar shape
for each texture group, but the level is different.
Avid
Roman
Gonzalez
Prof.
Dr. Mihai
Datcu
Competence Centre on Information Extraction
and Image Understanding for Earth Observation
Artifacts Detection Using Kolmogorov’s Structure
Function Approach
 To detect artifacts using Kolmogorov Structure Function
(KSF), the first step is to watch the behavior of the KSF
curve for images with artifacts and images without artifact.
 One aspect to consider is how to generate the candidates
for the necessary space S. For this purpose, we have
generated the candidates using 2 methods: Candidates
generation by JPEG lossy compression and Candidates
generation by genetic algorithm.
Avid
Roman
Gonzalez
Prof.
Dr. Mihai
Datcu
Competence Centre on Information Extraction
and Image Understanding for Earth Observation
2000
4
2.5
1800
x 10
1600
2
1400
1200
1.5
1000
800
1
600
400
0.5
200
0
0
500
1000
1500
2000
2500
3000
KSF using jpeg lossy compression
3500
0
0
0.5
1
1.5
2
KSF using genetic algorithm
2.5
 We can observe that the better discrimination is done when
we generate the candidates for the space S using the JPEG
lossy compression. Also using JPEG lossy compression the
approximation to the Rate-distortion analysis is better.
Avid
Roman
Gonzalez
Prof.
Dr. Mihai
Datcu
4
x 10
Competence Centre on Information Extraction
and Image Understanding for Earth Observation
 We use the jpeg lossy compression for generate candidates
and to draw the Kolmogorov Structure Function for each
patch of a satellite image and try to detect the artifacts. For
this experiments we use an image with aliasing introduces
manually.
Aliasing detection in city environmental using KSF and candidate generation with
JPEG lossy compression
Avid
Roman
Gonzalez
Prof.
Dr. Mihai
Datcu
Competence Centre on Information Extraction
and Image Understanding for Earth Observation
CONCLUSIONS
 The Kolmogorov structure function represents the
relationship between an element or data with its
model, structure, or explanation.
 In this work, we have used the Kolmogorov
structure function as a approximation of ratedistortion function using Kolmogorov complexity
theory and the complexity-distortion theory, so we
can examine the complexity of the images to be
analyzed, this complexity would be related to the
presence or absence of artifacts.
Avid
Roman
Gonzalez
Prof.
Dr. Mihai
Datcu
16/47
Competence Centre on Information Extraction
and Image Understanding for Earth Observation
CONCLUSIONS
 The generation of candidates for to calculation the
Kolmogorov structure function is an important step,
in this work was done experiments using 2
methods, generation of candidates by jpeg lossy
compression and candidate generation using
genetic algorithms, we obtain better results using
lossy jpeg compression.
Avid
Roman
Gonzalez
Prof.
Dr. Mihai
Datcu
17/47
Competence Centre on Information Extraction
and Image Understanding for Earth Observation
avid.roman-gonzalez@ieee.org
http://www.avid-romangonzalez.com
Avid
Roman
Gonzalez
Prof.
Dr. Mihai
Datcu