Catálogo de publicaciones - libros

EpiGauss is a method that combines single dipole model with dipole clustering to characterize active brain generators in space and time related to EEG events. EpiGauss was applied to study epileptogenic activity in 4 patients suffering of hypothalamic hamartoma related epilepsy, a rare syndrome with a unique epileptogenic source – the hamartoma lesion – and natural propagation hypothesis – from hamartoma to the surface EEG focus. The results are compared to Rap-MUSIC and Single Moving Dipole methods over the same patients.

This paper presents a methodology to automatically identify polygonal patterns on the surface of Mars. These structures, which are typical of periglacial regions, result from climate oscillations and present a wide variation in size, shape and topology and occur in different types of terrains with rather different constituents and spectral reflectances. The proposed approach is mainly based on the analysis of the dynamics of watershed contours and is successfully applied to a set of different types of patterned terrains of Mars shown by MGS/MOC images.

Methods for learning sparse classification are among the state-of-the-art in supervised learning. Sparsity, essential to achieve good generalization capabilities, can be enforced by using heavy tailed priors/ regularizers on the weights of the linear combination of functions. These priors/regularizers favour a few large weights and many to exactly zero. The Sparse Multinomial Logistic Regression algorithm [1] is one of such methods, that adopts a Laplacian prior to enforce sparseness. Its applicability to large datasets is still a delicate task from the computational point of view, sometimes even impossible to perform. This work implements an iterative procedure to calculate the weights of the decision function that is O ( m ^2) faster than the original method introduced in [1] ( m is the number of classes). The benchmark dataset Indian Pines is used to test this modification. Results over subsets of this dataset are presented and compared with others computed with support vector machines.

In this paper, we present a novel automatic approach for building detection from high resolution satellite image with densely build-up buildings. Unlike the previous approaches which normally start with lines and junctions, our approach is based on regions. In our method, first the prior building model is constructed with texture and shape features from the training building set. Then, we over-segment the input image into many small atomic regions. Given the prior building model and the over-segmented image, we group these small atomic regions together to generate region groups which have a similar pattern with the prior building model. These region groups are called candidate building region groups(CBRGs). The CBRGs grouping and recognition problems are formulated into an unified Bayesian probabilistic framework. In this framework, the CBRGs grouping and recognition are accomplished simultaneously by a stochastic Markov Chain Monte Carlo(MCMC) mechanism. To fasten this simulation process, an improved Swendsen-Wang Cuts graph partition algorithm are used. After obtaining CBRGs, lines which have strong relationship with CBRGs are extracted. From these lines and the CBRG boundaries, 2-D rooftop boundary hypotheses are generated. Finally, some contextual and geometrical rules are used to verify these rooftop boundary hypotheses. Experimental results are shown on areas with hundreds of buildings.

The striping noise removal method of an along-track scanned satellite image is considered in this paper. Nonuniformity of detectors caused by imperfect calibration and the drift of detector characteristics generates striping noise. The proposed nonlinear mapping consists of offset component correction (OCC) and nonlinear component correction (NCC). OCC is executed first under the assumption that the tendency of temporal (column) mean changes slowly across the detectors. Secondly, NCC, which is the least square approach for each of the same input intensity, is performed to reflect the nonlinear characteristics of the detector. The effectiveness of the proposed algorithm is demonstrated experimentally with real satellite images.

We analyze the capabilities of CASI data for the discrimination of vine varieties in hyperspectral images. To analyze the discrimination capabilities of the CASI data, principal components analysis and linear discriminant analysis methods are used. We assess the performance of various classification techniques: Multi-layer perceptrons, radial basis function neural networks, and support vector machines. We also discuss the trade-off between spatial and spectral resolutions in the framework of precision viticulture.

High resolution satellite images are now important data sources for map update in urban areas. Digital surface models (DSM) acquired by laser scanning are also becoming popular for urban planning and other applications. This paper deals with the integration of data sets of these types. An Ikonos image of an urban area was oriented, with sub-pixel accuracy, using the DSM as ground control, in order to produce a true ortho-image. This method works successfully, solving the important problem of relief displacement of buildings. A straightforward procedure of predicting shadow areas from the DSM was also attempted, in order to compensate for sha dow effects. The success was limited due to the time difference between the acquisition of the two datasets.

In this paper we present a new approach for the detection of defects in random colour textures. This approach is based on the use of the T^2 statistic and it is derived from the MIA strategy (Multivariate Image Analysis) developed in recent years in the field of applied statistics. PCA analysis is used to extract a reference eigenspace from a matrix built by unfolding the RGB raw data of defect-free images. The unfolding is performed compiling colour and spatial information of pixels. New testing images are also unfolded and projected onto the reference eigenspace obtaining a score matrix used to compute the T^2 images. These images are converted into defect maps which allow the location of defective pixels. Only very few samples are needed to perform unsupervised training. With regard to literature, the method uses one of the simplest approaches providing low computational costs.

Scene segmentation among background and foreground (moving) regions represents the first layer of many applications such as visual surveillance. Exploiting PTZ cameras permits to widen the field of view of a surveyed area and to achieve real object tracking through pan and tilt movements of the observer point of view. Having a mosaiced background allows a system to exploit the background subtraction technique even with moving cameras. Although spatial alignment issues have been thoroughly investigated, tonal registration has been often left out of consideration. This work presents a robust general purpose technique to perform spatial and tonal image registration to achieve a background mosaic without exploiting any prior information regarding the scene or the acquisition device. Accurate experiments accomplished on outdoor and indoor scenes assess the visual quality of the mosaic. Finally, the last experiment proves the effectiveness of using such a mosaic in our visual surveillance application.

In this paper, we propose a new approach of similarity retrieval in ancient handwritten documents. Similarities are retrieved on shape fragments which can be analysed according to different granularities (from the grapheme to the character size) depending on the handwriting quality, regularity and sharpness. The approach is based on a “segmentation free” methodology which considers the image with all frequencies and Gray levels. The method has been applied for two purposes: to establish the basis of a new handwritings compression approach based on handwritten images specificities and to show that only shape fragments can efficiently be used for writer characterization. We present here the global methodology for the similarities characterization which lies on an oriented handwriting shapes decomposition. First results for data compression and writer characterization are very encouraging in a field where there is no anterior work.