Iris Vanhamel

Multiscale gradient watersheds of color images

We present a new framework for the hierarchical segmentation of color images. The proposed scheme comprises a nonlinear scale-space with vector-valued gradient watersheds. Our aim is to produce a meaningful hierarchy among the objects in the image using three image components of distinct perceptual significance for a human observer, namely strong edges, smooth segments and detailed segments. The scale-space is based on a vector-valued diffusion that uses the Additive Operator Splitting numerical scheme. Furthermore, we introduce the principle of the dynamics of contours in scale-space that combines scale and contrast information. The performance of the proposed segmentation scheme is presented via experimental results obtained with a wide range of images including natural and artificial scenes.

A hierarchical Markovian model for multiscale region-based classification of vector-valued images

We propose a new classification method for vector-valued images, based on: 1) a causal Markovian model, defined on the hierarchy of a multiscale region adjacency tree (MRAT), and 2) a set of nonparametric dissimilarity measures that express the data likelihoods. The image classification is treated as a hierarchical labeling of the MRAT, using a finite set of interpretation labels (e.g., land cover classes). This is accomplished via a noniterative estimation of the modes of posterior marginals (MPM), inspired from existing approaches for Bayesian inference on the quadtree. The paper describes the main principles of our method and illustrates classification results on a set of artificial and remote sensing images, together with qualitative and quantitative comparisons with a variety of pixel-based techniques that follow the Bayesian-Markovian framework either on hierarchical structures or the original image lattice.

Scale Selection for Compact Scale-Space Representation of Vector-Valued Images

This paper investigates the scale selection problem for nonlinear diffusion scale-spaces. This topic comprises the notions of localization scale selection and scale space discretization. For the former, we present a new approach. It aims at maximizing the image content’s presence by finding the scale that has a maximum correlation with the noise-free image. For the latter, we propose to adapt the optimal diffusion stopping time criterion of Mrázek and Navara in such a way that it may identify multiple scales of importance.

Impact of Urban Land-Cover Classification on Groundwater Recharge Uncertainty

Objective and detailed mapping of urban land-cover types over large areas is important for hydrological modelling, as most man-made land-cover consist of sealed surfaces which strongly reduce groundwater recharge. Moreover, impervious surfaces are the predominant type in urbanized areas and can lead to increased surface runoff. Classification of man-made objects in urbanized areas is not straightforward due to similarity in spectral properties. This study examines the use of hyperspectral CHRIS-Proba images for complex urban land-cover classification of the Woluwe River catchment, Brussels, Belgium. Two methods are compared: 1) a multiscale region-based classification approach, which is based on a causal Markovian model being defined on a Multiscale Region Adjacency Tree and a set of nonparametric dissimilarity measures; and 2) a pixel based classification method with a Mahalanobis distance classifier. Multiscale region-based classification results in a Kappa value of 0.95 while pixel-based classification has a slightly lower Kappa value of 0.92. The impact of the classification method on the hydrology is estimated with the application of the WetSpass physically-based distributed water balance model. The model uncertainty is assessed with the use of a Monte Carlo simulation. Model results show that the region-based classification yields to a higher yearly recharge than the pixel-based classification. The overall uncertainty, quantified by the Monte Carlo method is lower for the region-based classification than for the pixel-based classification. The presented study indicates that the selection of the classification technique is of critical importance for the outcome of hydrological models.

Scale space segmentation of color images using watersheds and fuzzy region merging

A multi-resolution segmentation approach for color images is proposed. The scale space is generated using the Perona-Malik diffusion approach and the watershed algorithm is employed to produce the regions in each scale. The dynamics of contours and the relative entropy of color region distribution are estimated as region dissimilarity features across the scale-space stack, and combined using a fuzzy rule based system. A minima-linking process by downward projection is carried out and subsequently the region dissimilarity, combining color, scale and homogeneity is estimated for the finer scale (localization scale). The final segmentation is derived using a previously presented merging process. To validate its performance qualitative and quantitative results are provided.

Watershed-based multiscale segmentation method for color images using automated scale selection

An automated multiscale segmentation approach for color images is presented. The scale-space stack is generated us- ing the Perona-Malik diffusion approach and the watershed algo- rithm is employed to produce the regions at each scale. A minima- linking process by downward projection is carried out over the successive scales, and a region dissimilarity measure—combining scale, contrast, and homogeneity—is subsequently estimated on the finer scale (localization scale). The dissimilarity measure is es- timated as a function of two different features, i.e., the dynamics of contours and the relative entropy of color region distributions, com- bined by means of a fuzzy-rule-based system. A region-merging process is also applied to the localization scale to produce the final regions. To validate the performance of the proposed multiscale segmentation, qualitative and quantitative results are provided in comparison to its single-scale counterpart. We also deal with the topic of localization scale selection. This stage is critical for the final segmentation results and can be used as a preprocessing step for higher level computer vision applications as well. A preliminary study of localization scale selection techniques is carried out. A scale se- lection method that originates from the evolution of the probability distribution of a region homogeneity measure across the generated scales is proposed next. The proposed algorithm is finally compared to a previously reported approach to indicate its efficiency.

Automatic Watershed Segmentation of Color Images

This paper presents a fully automatic watershed color segmentation scheme which is an extension to color images of a previously reported approach dedicated to segmentation of scalar images. The importance of this extension lies mainly on its ability to automatically select an optimum result out of a hierarchical stack. This achievement is realized through the introduction of new evaluation methods for the segmentation quality of each level of the hierarchy which considers a tradeoff between the preservation of details and the suppression of heterogeneity. The first method estimates the local color error of the regions and combines it with the amount of regions. The second evaluates the contrast of the segmented image by combining a region uniformity with an inter-region contrast measure for all regions. These two methods are compared with respect to an existing one. Experimental results demonstrate the improvement which has been achieved by using the new evaluation criteria.

Hierarchical Segmentation Using Dynamics of Multiscale Color Gradient Watersheds

In this paper, we describe and compare two multiscale color segmentation schemes basedon the Gaussian multiscale and the Perona and Malik anisotropic diffusion. The proposed segmentation schemes consist of an extension to color images of an earlier multiscale hierarchical watershedsegm entation for scalar images. Our segmentation scheme constructs a hierarchy among the watershedre gions using the principle of dynamics of contours in scale-space. Each contour is valuated by combining the dynamics of contours over the successive scales. We conduct experiments on the scale-space stacks created by the Gaussian scale-space andt he Perona and Malik anisotropic diffusion scheme. Our experimental results consist of the comparison of both schemes with respect to the following aspects: size and in formation reduction between successive levels of the hierarchical stack, dynamics of contours in scale space and computation time.

Nonlinear Multiscale Graph Theory based Segmentation of Color Images

In this paper the issue of image segmentation within the framework of nonlinear multiscale watersheds in combination with graph theory based techniques is addressed. First, a graph is created which decomposes the image in scale and space using the concept of multiscale watersheds. In the subsequent step the obtained graph is partitioned using recursive graph cuts in a coarse to fine manner. In this way, we are able to combine scale and feature measures in a flexible way: the feature-set that is used to measure the dissimilarities may change as we progress in scale. We employ the earth movers distance on a featureset that combines color, scale and contrast features to measure the dissimilarity between the nodes in the graph. Experimental results demonstrate the efficiency of the proposed method for natural scene images

A hierarchical Markovian model for multiscale region-based classification of multispectral images

We propose a new multispectral image classification method, based on a Markovian model, defined on the hierarchy of a multiscale region adjacency graph. The paper describes the main principles of our method and illustrates classification results on a set of artificial and remote sensing images, together with qualitative and quantitative comparisons with a variety of multi-and single-resolution Bayesian classification approaches.