Related Experiment Video
Updated: May 14, 2026
06:25
Time Multiplexing Super Resolving Technique for Imaging from a Moving Platform
Published on: February 12, 2014
Minimum description length synthetic aperture radar image segmentation.
Frédéric Galland1, Nicolas Bertaux, Philippe Réfrégier
1Physics and Image Processing Group, Fresnel Institute UMR CNRS 6133, Ecole Nationale Supérieure de Physique de Marseille, Domaine universitaire de St Jerome, 13397 Marseille Cedex 20, France.
Summary
We introduce a new minimum description length (MDL) method using a deformable polygonal grid for automatic image segmentation. This approach effectively segments speckled images by optimizing grid complexity and probabilistic speckle properties.
Area of Science:
- Image Processing
- Computer Vision
- Statistical Modeling
Background:
- Speckle noise significantly degrades image quality in applications like Synthetic Aperture Radar (SAR).
- Accurate segmentation of homogeneous regions within speckled images is crucial for quantitative analysis.
Purpose of the Study:
- To develop a novel Minimum Description Length (MDL) approach for automatic image segmentation.
- To address the challenge of segmenting images corrupted by speckle noise.
Main Methods:
- A deformable polygonal grid is employed as a partition for image segmentation.
- The method minimizes a unique MDL criterion incorporating speckle properties and grid complexity.
- Noise parameters are estimated using maximum likelihood-like methods.
Main Results:
- The proposed MDL approach enables automatic estimation of the number of regions, nodes, and node locations.
- The method achieves a global MDL criterion without undetermined parameters.
- Performance is validated on synthetic and real SAR agricultural images.
Conclusions:
- The deformable partition MDL approach provides an effective and parameter-free solution for speckled image segmentation.
- This technique offers robust segmentation by considering image statistics and model complexity.
- The study demonstrates the utility of MDL for advanced image analysis tasks.
