Texture features have proved to be an important tool in image segmentation and object recognition, as well as interpretation of images in a variety of applications ranging from medical imaging to remote sensing. Many methods were suggested to achieve good discrimination between different textural regions. We propose a non-supervised classification method. The method combines a multi-resolution based texture feature with features based on first and second order statistics. These features are calculated for each pixel in the image and its neighbours. A clustering algorithm, based on the generalized Lloyd algorithm, is applied and finally an iterative region merging process, based on the phagocytes heuristic, is used to improve the classification of the borders of the regions and to reduce local clustering errors. Examples of the method and the features effectiveness are presented using SPOT satellite images.

Texture classification in aerial photographs and satellite data

A nonparametric approach to linear feature extraction is presented. The theoretical background is introduced with a derivation of the equation that gives the best scalar extractor according to Patrick-Fischer distance. The outlines of the implementation are given. The method is applied to the classification of binary synthetic textures with natural visual aspect. The performances of the proposed method are shown to be better than the Fisher discriminant-analysis-based classifier. Concluding remarks are given for future improvements, further applications, and theoretical discussion. >

http://ieeexplore.ieee.org/iel2/220/1164/00028433.pdf

A nonparametric approach to linear feature extraction; application to classification of binary synthetic textures

Textural analysis is now a commonly used technique in digital image processing. In this paper, we present an application of textural analysis to high resolution SPOT satellite images. The purpose of the methodology is to improve classification results, i.e. image segmentation in remote sensing. Remote sensing techniques, based on high resolution satellite data offer good perspectives for the cartography of littoral environment. Textural information contained in the pan-chromatic channel of ten meters resolution is introduced in order to separate different types of structures. The technique we used is based on statistical pattern recognition models and operates in two steps. A first step, features extraction, is derived by using a stepwise algorithm. Segmentation is then performed by cluster analysis using these extracted. features. The texture features are computed over the immediate neighborhood of the pixel using two methods : the cooccurence matrices method and the grey level difference statistics method. Image segmentation based only on texture features is then performed by pixel classification and finally discussed. In a future paper, we intend to compare the results with aerial data in view of the management of the littoral resources.

A. Hillion

Papers

Image Segmentation By Cluster Analysis Of High Resolution Textured SPOT Images