Catálogo de publicaciones - libros

The paper focusses on a group of segmentation problems dealing with 3D data sets showing thin objects that appear disconnected in the data due to partial volume effects or a large spacing between neighbouring slices. We propose a modification of the speed function for the well-known level set method to bridge these discontinuities. This allows for the segmentation of the object as a whole. In this paper we are concerned with treelike structures, particularly dendrites in microscopic data sets, whose shape is unknown prior to segmentation. Using the modified speed function, our algorithm segments dendrites and their spines, even if parts of the object appear to be disconnected due to artifacts.

Due to its broad impact in many image analysis applications, the problem of image segmentation has been widely studied. However, there still does not exist any automatic segmentation procedure able to deal accurately with any kind of image. Thus semi-automatic segmentation methods may be seen as an appropriate alternative to solve the segmentation problem. Among these methods, the marker-based watershed has been successfully involved in various domains. In this algorithm, the user may locate the markers, which are used only as the initial starting positions of the regions to be segmented. We propose to base the segmentation process also on the contents of the markers through a supervised pixel classification, thus resulting in a knowledge-based watershed segmentation where the knowledge is built from the markers. Our contribution has been evaluated through some comparative tests with some state-of-the-art methods on the well-known Berkeley Segmentation Dataset.

This paper proposes a model-based method for intensity-based segmentation of images acquired from multiple modalities. Pixel intensity within a modality image is represented by a univariate Gaussian distribution mixture in which the components correspond to different segments. The proposed Multi-Modality Expectation-Maximization () algorithm then estimates the probability of each segment along with parameters of the Gaussian distributions for each modality by maximum likelihood using the Expectation-Maximization (EM) algorithm. Multimodal images are simultaneously involved in the iterative parameter estimation step. Pixel classes are determined by maximising a probability contributed from all multimodal images. Experimental results show that the method exploits and fuses complementary information of multimodal images. Segmentation can thus be more precise than when using single-modality images.

In this paper we introduce a new method for image segmentation. It is based on a height map generated from the input image. The height map characterizes the image content in such a way that the application of the watershed concept provides a proper segmentation of the image. The height map enables the watershed method to provide better segmentation results on difficult images, e.g., images of natural objects, than without the intermediate height map generation. Markers used for the watershed concept are generated automatically from the input data holding the advantage of a more autonomous segmentation. In addition, we introduce a new edge detector which has some advantages over the Canny edge detector. We demonstrate our methods by means of a number of segmentation examples.

An orientability measure determines how orientable a shape is; i.e. how reliable an estimate of its orientation is likely to be. This is valuable since many methods for computing orientation fail for certain shapes. In this paper several existing orientability measures are discussed and several new orientability measures are introduced. The measures are compared and tested on synthetic and real data.

Thinning is an iterative layer by layer erosion for extracting skeleton. This paper presents an efficient parallel 3D thinning algorithm which produces medial surfaces. A three–subiteration strategy is proposed: the thinning operation is changed from iteration to iteration with a period of three according to the three deletion directions.

Active Shape Models often require a considerable number of training samples and landmark points on each sample, in order to be efficient in practice. We introduce the Fractal Active Shape Models, an extension of Active Shape Models using fractal interpolation, in order to surmount these limitations. They require a considerably smaller number of landmark points to be determined and a smaller number of variables for describing a shape, especially for irregular ones. Moreover, they are shown to be efficient when few training samples are available.

The eccentricity transform associates to each point of a shape the shortest distance to the point farthest away from it. It is defined in any dimension, for open and closed manyfolds. Top-down decomposition of the shape can be used to speed up the computation, with some partitions being better suited than others. We study basic convex shapes and their decomposition in the context of the continuous eccentricity transform. We show that these shapes can be decomposed for a more efficient computation. In particular, we provide a study regarding possible decompositions and their properties for the ellipse, the rectangle, and a class of elongated shapes.

Ellipse is one of the most common features that appears in images. Over years in research, real-timing and robustness have been two very challenging problems aspects of ellipse detection. Aiming to tackle them both, we propose an ellipse detection algorithm based on pseudo-random sample consensus (PRANSAC). In PRANSAC we improve a contour-based ellipse detection algorithm (CBED), which was presented in our previous work. In addition, the parallel thinning algorithm is employed to eliminate useless feature points, which increases the time efficiency of our detection algorithm. In order to further speed up, a 3-point ellipse fitting method is introduced. In terms of robustness, a “robust candidate sequence” is proposed to improve the robustness performance of our detection algorithm. Compared with the state-of-the-art ellipse detection algorithms, experimental results based on real application images show that significant improvements in time efficiency and performance robustness of the proposed algorithm have been achieved.

In this paper we introduce a new method for computing the orientation for compound shapes. If the method is applied to single component shapes the computed orientation is consistent with the shape orientation defined by the axis of the least second moment of inertia. If the new method is applied to compound shapes this is not the case, and consequently the presented method is both new and different.