5 igarss2011_mdm.pdf
TRANSCRIPT
A General Approach to the Spatial Simpli�cation of Remote Sensing
Images Based on Morphological Connected Filters
Mauro Dalla Mura†,�, Jón Atli Benediktsson�, Lorenzo Bruzzone†
†Department of Information Engineering and Computer ScienceUniversity of Trento.
�Faculty of Electrical and Computer EngineeringUniversity of Iceland.
IGARSS 2011
24-29 July
Outline
1 Introduction
2 General Approach for Image Simpli�cation
Connected Operators
Methodology
3 Conclusion and Future Developments
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 2 / 19
Introduction
Remote Sensing VHR Images
QuickBird 60cm, Panchromatic image, Bam (Iraq) Geoeye 50cm, Pansharpened images, Vancouver (Canada)c©Google
ROSIS-03 1.3m, Hyperspectral image, Pavia (Italy) TerraSAR-X 1.1m, Spotlight SAR, Dorsten (Germany)
The information extraction in remote sensing images is becoming increasingly
complex due to the progressively higher spatial resolution of the current
sensors.
How to extract the informative components dealing with the huge amount
of details?
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 3 / 19
Introduction
Remote Sensing VHR Images
QuickBird 60cm, Panchromatic image, Bam (Iraq) Geoeye 50cm, Pansharpened images, Vancouver (Canada)c©Google
ROSIS-03 1.3m, Hyperspectral image, Pavia (Italy) TerraSAR-X 1.1m, Spotlight SAR, Dorsten (Germany)
The information extraction in remote sensing images is becoming increasingly
complex due to the progressively higher spatial resolution of the current
sensors.
How to extract the informative components dealing with the huge amount
of details?
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 3 / 19
Introduction
Image Simpli�cation
Spatial simplication of the image.
Pre-processing operation for many remote sensing applications:I Image segmentation;I Supervised/unsupervised thematic classi�cation;I Land cover change analysis;I Object recognition and extraction;I Denoising SAR images;I Analysis of multiangular images.
Image simpli�cation leads to:I noise reduction;I exploiting the contextual relations;I modeling spatial relations;I removing or attenuating undesired details.
Simpli�cation of the image performed by spatial �ltering, a 2-step procedurecomposed of:
1 selection of the �lters parameters;2 application of the operator on the image.
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 4 / 19
Introduction
Image Simpli�cation (examples)
VHR optical image - di�erent simpli�cations
Which details should be removed?
Which operator should be applied? And which �lter parameters should be
selected?
It depends on the application and on the type of image.
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 5 / 19
Introduction
Image Simpli�cation (examples)
VHR optical image - di�erent simpli�cations
Which details should be removed?
Which operator should be applied? And which �lter parameters should be
selected?
It depends on the application and on the type of image.
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 5 / 19
Introduction
Image Simpli�cation (examples)
VHR optical image - di�erent simpli�cations
Which details should be removed?
Which operator should be applied? And which �lter parameters should be
selected?
It depends on the application and on the type of image.
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 5 / 19
Introduction
Motivation
Issues related to image simpli�cation
The selection of the parameters of the �lters is application dependent.
Proper operators should be used.
Manual operation.
Aims of the work
De�ne a novel general approach to image simpli�cation
based on morphological connected operators;
suitable for the processing of di�erent types of images and di�erent
applications;
suitable to be performed in an automated way.
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 6 / 19
General Approach for Image Simpli�cation Connected Operators
Connected Operators - Operators by Reconstruction
Connected operators are morphological �lters that process an image by only
merging its �at zones.
Either completely remove or entirely preserve a region in the image.
They do not distort the geometrical characteristics (e.g., shape, edges) of the
structures in the image.
Operators by Reconstruction
Closing
φB(f )
Closing by rec.
φBR(f ) = Rε
f[δB(f )]
Original image
f
Opening by rec.
γBR(f ) = Rδ
f[εB(f )]
Opening
γB(f )
Connected operators are suitable for the analysis of VHR images.
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 7 / 19
General Approach for Image Simpli�cation Connected Operators
Connected Operators - Operators by Reconstruction
Connected operators are morphological �lters that process an image by only
merging its �at zones.
Either completely remove or entirely preserve a region in the image.
They do not distort the geometrical characteristics (e.g., shape, edges) of the
structures in the image.
Operators by Reconstruction
Closing
φB(f )
Closing by rec.
φBR(f ) = Rε
f[δB(f )]
Original image
f
Opening by rec.
γBR(f ) = Rδ
f[εB(f )]
Opening
γB(f )
Connected operators are suitable for the analysis of VHR images.
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 7 / 19
General Approach for Image Simpli�cation Connected Operators
Connected Operators - Attribute Filters
Attribute �lters are connected operators de�ned in the mathematical
morphology framework and recently used for the analysis of remote sensing1.
Based on a measure (attribute) computed on the regions of an image.
Filtering performed by removing the regions that do not ful�ll a condition (T )
which compares an attribute attr against a reference value λ
(e.g., T = attr ≥ λ).Main operators:I attribute thinning, γT ;I attribute thickening, φT .
Example: Di�erent attributes
Original image Area Standard deviation Moment of inertia Solidity
1 M. Dalla Mura, J. A. Benediktsson, B. Waske, and L. Bruzzone, �Morphological attribute pro�les for the analysisof very high resolution images,� IEEE Transactions on Geoscience and Remote Sensing, vol. 48, no. 10, pp.3747 � 3762, Oct. 2010.IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 8 / 19
General Approach for Image Simpli�cation Methodology
Architecture of the Proposed General Approach
Scenario 3
ApplicationKnowledge
XParameters Selection
Y
SceneKnowledge
Scenario 1
Scenario 2
Bank ofConnected Filters
e. g., attribute thinning,attribute thickening, ...
PerformanceAssessment
The proposed approach is composed of two modules that perform the operations
of:
1 selection of the parameters and operators;
2 �ltering.
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 9 / 19
General Approach for Image Simpli�cation Methodology
Architecture of the Proposed General Approach
Scenario 3
ApplicationKnowledge
XParameters Selection
Y
SceneKnowledge
Scenario 1
Scenario 2
Bank ofConnected Filters
e. g., attribute thinning,attribute thickening, ...
PerformanceAssessment
The proposed approach is composed of two modules that perform the operations
of:
1 selection of the parameters and operators;
2 �ltering.
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 9 / 19
General Approach for Image Simpli�cation Methodology
Architecture of the Proposed General Approach
Scenario 3
ApplicationKnowledge
XParameters Selection
Y
SceneKnowledge
Scenario 1
Scenario 2
Bank ofConnected Filters
e. g., attribute thinning,attribute thickening, ...
PerformanceAssessment
The proposed approach is composed of two modules that perform the operations
of:
1 selection of the parameters and operators;
2 �ltering.
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 9 / 19
General Approach for Image Simpli�cation Methodology
Filter Selection and Operative Scenarios
Operative Scenarios
Information available as prior
knowledge:
1 Scenario 1Scene Knowledge %
Application Knowledge %
2 Scenario 2Scene Knowledge %
Application Knowledge "
3 Scenario 3Scene Knowledge "
Application Knowledge "
Scenario 3
ApplicationKnowledge
XParameters Selection
Y
SceneKnowledge
Scenario 1
Scenario 2
Bank ofConnected Filters
e. g., attribute thinning,attribute thickening, ...
PerformanceAssessment
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 10 / 19
General Approach for Image Simpli�cation Methodology
Filter Selection and Operative Scenarios
Operative Scenarios
Information available as prior
knowledge:
1 Scenario 1Scene Knowledge %
Application Knowledge %
2 Scenario 2Scene Knowledge %
Application Knowledge "
3 Scenario 3Scene Knowledge "
Application Knowledge "
Scenario 3
ApplicationKnowledge
XParameters Selection
Y
SceneKnowledge
Scenario 1
Scenario 2
Bank ofConnected Filters
e. g., attribute thinning,attribute thickening, ...
PerformanceAssessment
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 10 / 19
General Approach for Image Simpli�cation Methodology
Filter Selection and Operative Scenarios
Operative Scenarios
Information available as prior
knowledge:
1 Scenario 1Scene Knowledge %
Application Knowledge %
2 Scenario 2Scene Knowledge %
Application Knowledge "
3 Scenario 3Scene Knowledge "
Application Knowledge "
Scenario 3
ApplicationKnowledge
XParameters Selection
Y
SceneKnowledge
Scenario 1
Scenario 2
Bank ofConnected Filters
e. g., attribute thinning,attribute thickening, ...
PerformanceAssessment
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 10 / 19
General Approach for Image Simpli�cation Methodology
Data set
Panchromatic image
Quickbird
panchromatic
image of 995×995pixels, 0.6 m
resolution.
Acquired over a
residential urban
area of Bam, Iran.
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 11 / 19
General Approach for Image Simpli�cation Methodology
Scenario 1 - Application Knowledge %, Scene Knowledge %
Aim: Generic reduction of the complexity of the image by reducing non
informative components.
Filtering aiming at reducing:
1 Noisy components ⇒ Removing small regions with values signi�cantly di�erent
from their surroundings;2 Inter-object variability ⇒ Flattening small values di�erences in homogeneous
regions.
Suitable to cope with most of the applications.
Eases the interpretation of the scene.
Exploits the contextual relations of the pixels.
Fully automatic suitable for batch processing.
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 12 / 19
General Approach for Image Simpli�cation Methodology
Scenario 1 - Example
Generic simpli�cation performed by:
γTφT with area attribute (remove small bright and dark regions);
γT with T based on relations between the regions1 (merge nested regions).
VHR image
Building rooftop (80×60 pixels). 2789 �at regions. Simpli�ed image. 1059 regions.
1 V. Caselles and P. Monasse, Geometric Description of Images as Topographic Maps. Springer, 2010.
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 13 / 19
General Approach for Image Simpli�cation Methodology
Scenario 2 - Application Knowledge ", Scene Knowledge %
Filters parameters selected according to the application.
The translation of the characteristics of the objects of interest from the
concept to the �lter parameters.
Example:I Application: building extraction.I Aim of the simpli�cation: enhance rectangular regionsI Concept: �keep rectangular regions�.I Filtering: attribute �lter with criterion: {rectangularity > 0.5});
Automation
Modeling the range of values of the features that drive the �lters with fuzzy
possibilistic functions.
Defuzzify in order to get the values for the �lters parameters.
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 14 / 19
General Approach for Image Simpli�cation Methodology
Scenario 2 - Enhancement of Buildings (Particulars)
Panchromatic image (f )
Filter by rec. f − γBR(f ) (B: disk radius 7 pixels)
Attribute �lter γT (f ) with T = (R > 0.3) ∧ (I < 0.5) ∧ (50 < A < 5000)R: rectangularity; I: moment of inertia; A: area
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 15 / 19
General Approach for Image Simpli�cation Methodology
Scenario 2 - Enhancement of Buildings (Particulars)
Panchromatic image (f )
Filter by rec. f − γBR(f ) (B: disk radius 11 pixels)
Attribute �lter γT (f ) with T = (R > 0.3) ∧ (I < 0.5) ∧ (50 < A < 5000)R: rectangularity; I: moment of inertia; A: area
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 15 / 19
General Approach for Image Simpli�cation Methodology
Scenario 2 - Enhancement of Buildings (Particulars)
Panchromatic image (f )
Filter by rec. f − γBR(f ) (B: disk radius 15 pixels)
Attribute �lter γT (f ) with T = (R > 0.3) ∧ (I < 0.5) ∧ (50 < A < 5000)R: rectangularity; I: moment of inertia; A: area
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 15 / 19
General Approach for Image Simpli�cation Methodology
Scenario 2 - Enhancement of Buildings (Particulars)
Panchromatic image (f )
Filter by rec. f − γBR(f ) (B: disk radius 19 pixels)
Attribute �lter γT (f ) with T = (R > 0.3) ∧ (I < 0.5) ∧ (50 < A < 5000)R: rectangularity; I: moment of inertia; A: area
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 15 / 19
General Approach for Image Simpli�cation Methodology
Scenario 2 - Enhancement of Dark Elongated Structures (Particulars)
Panchromatic image (f )
Filter by rec. C[φBR(f )− f ] (B: disk radius 3 pixels)
Attribute �lter φT (f ) with T = (H > 10000) ∧ (I > 1.0)H: height; I: moment of inertia
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 16 / 19
General Approach for Image Simpli�cation Methodology
Scenario 2 - Enhancement of Dark Elongated Structures (Particulars)
Panchromatic image (f )
Filter by rec. C[φBR(f )− f ] (B: disk radius 7 pixels)
Attribute �lter φT (f ) with T = (H > 10000) ∧ (I > 1.0)H: height; I: moment of inertia
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 16 / 19
General Approach for Image Simpli�cation Methodology
Scenario 2 - Enhancement of Dark Elongated Structures (Particulars)
Panchromatic image (f )
Filter by rec. C[φBR(f )− f ] (B: disk radius 11 pixels)
Attribute �lter φT (f ) with T = (H > 10000) ∧ (I > 1.0)H: height; I: moment of inertia
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 16 / 19
General Approach for Image Simpli�cation Methodology
Scenario 2 - Enhancement of Dark Elongated Structures (Particulars)
Panchromatic image (f )
Filter by rec. C[φBR(f )− f ] (B: disk radius 15 pixels)
Attribute �lter φT (f ) with T = (H > 10000) ∧ (I > 1.0)H: height; I: moment of inertia
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 16 / 19
General Approach for Image Simpli�cation Methodology
Scenario 3 - Application Knowledge ", Scene Knowledge "
Ad hoc parameters selection. If the available information is a set of labeled
samples (i.e., a training set), the reduction of the image complexity
generated by the �ltering aims at increasing the separability of the classes.
Performance assessment. The quality of the simpli�cation obtained can be
evaluated on the known samples according to a given criterion.
Automation
De�ne a cost function to minimize, representing the �tness of the generated
result with the input requirements;
De�ne a optimization procedure and a stopping condition.
See on Thurstday: S. Peeters, P. R. Marpu, J. A. Benediktsson, M. Dalla Mura �Classi�cation using extended
morphological attribute pro�les based on di�erent feature extraction techniques.�
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 17 / 19
Conclusion and Future Developments
Conclusion and Future Developments
Conclusion
De�nition of a novel general approach for image simpli�cation based on
connected �lters (in particular attribute �lters).Suitable forI processing many di�erent image types;I di�erent applications involving image analysis.
Contributions:I identi�cation of three scenarios modeling common di�erent operative
conditions;I giving guidelines for the automation of the process;I qualitative evaluation of the proposed approach on a real data set in di�erent
scenarios.
Future Developments
Extensively test the approach on di�erent type of images and applications.
Improve the automation of the process.
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 18 / 19
Conclusion and Future Developments
Thanks for your attention!
IGARSS 2011 (24-29 July) Mauro Dalla Mura [email protected] 19 / 19