ANALYSE COMPARATIVE DE LA CLASSIFICATION CONTEXTUELLE ET DU MAXIMUM DE VRAISEMBLANCE : SYNTHÈSE ET CAS D'ÉTUDE

* Correspondant principal Les images a tres haute resolution se caracterisent par une variance locale elevee, rendant peu efficaces les classificateurs conventionnels tels que le maximum de vraisemblance. Les classificateurs dits contextuels paraissent mieux adaptes au traitement de ce type d'images. Dans cet article, on presente une revue des principes de la classification contextuelle et, par une classification d'une image aerienne, une illustration des comportements respectifs des classificateurs de maximum de vraisemblance et de « croissance de region » (logiciel eCognition). L'image aerienne utilisee, d’une resolution spatiale de 0,5 m, contient des classes d'occupation du sol reputees pour presenter des signatures spectrales peu disjointes. Le comportement respectif des deux classificateurs est bien mis en evidence, notamment l'avantage offert par le classificateur contextuel d'imposer, pour l'agregation de nouveaux pixels, la double condition d'une contiguite spatiale et d'une similarite spectrale.

[1]  Roland Wilson,et al.  Unsupervised image segmentation combining region and boundary estimation , 2000, Image Vis. Comput..

[2]  Giuseppe Arbia,et al.  Contextual classification in image analysis: an assessment of accuracy of ICM , 1999 .

[3]  Luc Vincent,et al.  Watersheds in Digital Spaces: An Efficient Algorithm Based on Immersion Simulations , 1991, IEEE Trans. Pattern Anal. Mach. Intell..

[4]  B. Laget,et al.  Color image segmentation using region adjacency graphs , 1997 .

[5]  M. Nagao,et al.  Edge preserving smoothing , 1979 .

[6]  B. Guindon,et al.  A Framework for the Development and Assessment of Object Recognition Modules from High-Resolution Satellite Images , 2000 .

[7]  B. Guindon.  Application of Perceptual Grouping Concepts to the Recognition of Residential Buildings in High Resolution Satellite Images , 2001 .

[8]  Kannan,et al.  ON IMAGE SEGMENTATION TECHNIQUES , 2022 .

[9]  Jaakko Astola,et al.  Clustering based region growing algorithm for color image segmentation , 1997, Proceedings of 13th International Conference on Digital Signal Processing.

[10]  Cristina Urdiales,et al.  Scale-dependent hierarchical unsupervised segmentation of textured images , 2001, Pattern Recognit. Lett..

[11]  Josef Kittler,et al.  Region growing: a new approach , 1998, IEEE Trans. Image Process..

[12]  James C. Tilton,et al.  Image segmentation by region growing and spectral clustering with a natural convergence criterion , 1998, IGARSS '98. Sensing and Managing the Environment. 1998 IEEE International Geoscience and Remote Sensing. Symposium Proceedings. (Cat. No.98CH36174).

[13]  B. Guindon,et al.  Computer-Based Aerial Image Understanding: A Review and Assessment of its Application to Planimetric Information Extraction from Very High Resolution Satellite Images , 1997 .

[14]  Xavier Cufí,et al.  A concurrent region growing algorithm guided by circumscribed contours , 2000, Proceedings 15th International Conference on Pattern Recognition. ICPR-2000.

[15]  Milan Sonka,et al.  Image Processing, Analysis and Machine Vision , 1993, Springer US.

[16]  Jianping Fan,et al.  Automatic image segmentation by integrating color-edge extraction and seeded region growing , 2001, IEEE Trans. Image Process..

[17]  I. V. Muralikrishna,et al.  Application of region growing algorithm for classification of IRS multispectral image , 1999, IEEE 1999 International Geoscience and Remote Sensing Symposium. IGARSS'99 (Cat. No.99CH36293).

[18]  P. Bolon,et al.  Analyse d'images: filtrage et segmentation , 1995 .