Mohamed Abadi

Texture features and segmentation based on multifractal approach

In this paper, we use a multifractal approach based on the computation of two spectrums for image analysis and texture segmentation problems. The two spectrums are the Legendre Spectrum, determined by classical methods, and the Large Deviation Spectrum, determined by kernel density estimation. We propose a way for the fusion of these two spectrums to improve textured image segmentation results. An unsupervised k-means is used as clustering approach for the texture classification. The algorithm is applied on mosaic image built using IKONOS images and various natural textures from the Brodatz album. The segmentation obtained with our approach gives better results than the application of each spectrum separately.

Parametric models of linear prediction error distribution for color texture and satellite image segmentation

In this article we present a Bayesian color texture segmentation framework based on the multichannel linear prediction error. Two-dimensional causal and non-causal real (in RGB color space) and complex (in IHLS and L^*a^*b^* color spaces) multichannel linear prediction models are used to characterize the spatial structures in color images. The main contribution of this segmentation methodology resides in the robust parametric approximations proposed for the multichannel linear prediction error distribution. These are composed of a unimodal approximation based on the Wishart distribution and a multimodal approximation based on the multivariate Gaussian mixture models. For the spatial regularization of the initial class label estimates, computed through the proposed parametric priors, we compare the conventional Potts model to a Potts model fusioned with a region size energy term. We provide performances of the method when using Iterated Conditional Modes algorithm and simulated annealing. Experimental results for the segmentation of synthetic color textures as well as high resolution QuickBird and IKONOS satellite images validate the application of this approach for highly textured images. Advantages of using these priors instead of classical Gaussian approximation and improved label field model are shown by these results. They also verify that the L^*a^*b^* color space exhibits better performance among the used color spaces, indicating its significance for the characterization of color textures through this approach.

Legendre Spectrum for texture classification

This paper deals with texture classification using a multifractal approach. More precisely we analyse the singularity/regularity exponent that composes the textures as they theoretically carry most of the information. The analysis is made using the Legendre spectrum. Then a parameter vector is computed to describe this spectrum in order to classify the textures with an unsupervised k-means classifier. The resulting algorithm is evaluated against a classification directly based on the textures

Grassland species characterization for plant family discrimination by image processing

Pasture species belonging to poaceae and fabaceae families constitute of essential elements to maintain natural and cultivated regions. Their balance and productivity are key factors for good functioning of the grassland ecosystems. The study is based on a process of image processing. First of all an individual signature is defined while considering geometric characteristics of each family. Then, this signature is used to discriminate between these families. Our approach focuses on the use of shape features in different situations. Specifically, the approach is based on cutting the representative leaves of each plant family. After cutting, we obtain leaves sections of different sizes and random geometry. Then, the shape features are calculated. Principal component analysis is used to select the most discriminatory features. The results will be used to optimize the acquisition conditions. We have a discrimination rate of more than 90% for the experiments carried out in a controlled environment. Experiments are being carried out to extend this study in natural environments.

Information criteria performance for feature selection

This paper shows the information criteria (IC) performances in feature selection framework. Feature selection aims to select a representative subset among a wide set of features. We apply this approach to classify an hand segmented image. The performance is tested using various feature selection schemes (SFS, SBS, SFFS and SBFS) to select the candidate subsets. The accuracy of the approach is based on a good quality of the joint probability density approximation of the combined features. They are obtained using histogram optimized thanks to the adaptive arithmetic coding principles. Our approach is tested on different reference data. The subsets quality is evaluated using correct classification rate computed on multiple classifiers. Results show stability and convergence properties of this tool and its ability to select representative subsets (in the sense that the subset of feature is a good characterization of the classes in which the data belong). Information Criteria could be used for feature selection as a good alternative to other criteria.

Hybrid Color Space Choice: An Optimisation Review for Cost/Efficiency Trade-Off

This paper deals with image representation improvement using hybrid color spaces. This representation is important because it influences segmentation and classification results. We present two improvements of an existing supervised algorithm to obtain the most adapted hybrid color space for a given image. These improvements are based on a multi-objective optimization leading to a cost-efficiency trade-off, and have a theoretical justification. A comparison of the different approaches shows that the most adapted hybrid color space is reached with our algorithm and improves classification results.

A Pareto Front Approach for Feature Selection

This article deals with the multi-objective aspect of an hybrid algorithm that we propose to solve the feature subset selection problem. The hybrid aspect is due to the sequence of a filter and a wrapper method. The filter method reduces the exploration space by keeping subsets having good internal properties and the wrapper method chooses among the remaining subsets with a classification performances criterion. In the filter step, the subsets are evaluated in a multi-objective way to ensure diversity within the subsets. The evaluation is based on the mutual information to estimate the dependency between features and classes and the redundancy between features within the same subset. We kept the non-dominated (Pareto optimal) subsets for the second step. In the wrapper step, the selection is made according to the stability of the subsets regarding classification performances during learning stage on a set of classifiers to avoid the specialization of the selected subsets for a given classifiers. The proposed hybrid approach is experimented on a variety of reference data sets and compared to the classical feature selection methods FSDD and mRMR. The resulting algorithm outperforms these algorithms.