Catálogo de publicaciones - libros

In this paper a novel approach for detecting fingerprint regions with relevant information is presented. This method is based on the capability of the wavelet transform to select image information considering at the same time spatial and frequency domains. The method has been tested with two fingerprint data bases providing excellent results. With this method the fingerprint core can be detected and also the background can be detached, providing an efficient region selection for any feature extraction method, preprocessing and matching algorithms.

In previous work [5] we have identified the gradient of the surface as the best representation for constructing Cartesian models of faces. This representation proved capable of capturing variations in facial shape over a sample of training data. The resulting statistical model can be fitted to Lambertian data using a simple non-exhaustive parameter adjustment procedure. In this paper we test the ability of the surface gradient-based model in two directions. First, we deal with non-lambertian images. Second, we use the model for face recognition purposes. Experiments with real world images suggest that the surface gradient model with the proposed parameter search can be used for accurate face shape recovery, showing a potential for face recognition applications.

This paper describes a technique for the representation of the 2D frontal view of faces, based on Catmull-Rom splines. It takes advantage of the a priori knowledge about the face structure and of the proprieties of Catmull-Rom splines, like interpolation, smoothness and local control, in order to define a set of key points that correspond among different faces. Moreover, it can compactly describe the whole face even if the face features have not been completely localized. The proposed model has been tested in practical contexts of face analysis and promising qualitative results are included to illustrate its versatility and accuracy.

In this paper we present a methodology to automatically analyze the soft tissue of the mouth. The methodology is based on some measures obtained from the lip contour. Process starts with the face localization, followed by the mouth area segmentation. A first approximation to the external lip contour is obtained by using active contours. The control points of the contours are used to calculate the four parametric functions that define the mouth template. Results show how the feature extraction algorithms and active contour adjustment perform. In addition, some tests were carried out on images of children with repaired cleft lip.

The use of 2D colour adjacency histograms for image matching in image retrieval scenarios is investigated. We present an algorithm for extracting representative colours from an image and a new method for matching 1D colour histograms and 2D colour adjacency histograms obtained from images quantised using different colour palettes. An experimental evaluation of the matching performance is done.

In this paper, we present a novel algorithm to detect homogeneous color regions in images. We show its performance by applying it to skin detection. In contrast to previously presented methods, we use only a rough skin direction vector instead of a static skin model as a priori knowledge. Thus, higher robustness is achieved in images captured under unconstrained conditions. We formulate the segmentation as a clustering problem in color space. A homogeneous color region in image space is modeled using a 3D gaussian distribution. Parameters of the gaussians are estimated using the EM algorithm with spatial constraints. We transform the image by a whitening transform and then apply a fuzzy k-means algorithm to the hue value in order to obtain initialization parameters for the EM algorithm. A divisive hierarchical approach is used to determine the number of clusters. The stopping criterion for further subdivision is based on the edge image.

For evaluation, the proposed method is applied to skin segmentation and compared with a well known method.

The human visual system is able to determine the color of objects irrespective of the power distribution illuminating the scene. This ability is called color constancy. It would be highly interesting to mimic this ability of the human visual system. Accurate color reproduction is very important from consumer photography to automatic color based object recognition. In theory, if we knew the color of the illuminant for each image pixel, we would be able to compute a color corrected image which is independent of the illuminant. We suggest the use of anisotropic diffusion to estimate the illuminant locally for each image pixel.

We propose an algorithm for restoration from multiple color images degraded by camera motion blur. We consider the special case when the camera moves in one plane perpendicular to the optical axis without any rotations. The algorithm needs to know neither camera motion nor camera parameters. The proposed algorithm belongs to the group of variational methods that estimate simultaneously sharp image and depth map, based on the minimization of a cost functional. Feasibility of the algorithm is demonstrated by two experiments with real images.

This paper proposes a real-time method for the detection and elimination of brightness in color images. We use a 2D-histogram that allows us to relate the signals of luminance and saturation of a color image and to identify the specularities in a given area of the histogram. This is known as the MS diagram and it is constructed from a polar color model. We use a new connected vectorial filter based on color morphology to eliminate the brightness. This filter operates only in the bright zones previously detected, reducing the high cost of processing of connected filtersand avoiding over-simplification, in single-processing and multiprocessing environments.

This paper presents a novel shape recovery technique that combines photometric stereo with polarization information. First, a set of ambiguous surface normals are estimated from polarization data. This is achieved using Fresnel theory to interpret the polarization patterns of light reflected from dielectric surfaces. The process is repeated using three different known light source positions. Photometric stereo is then used to disambiguate the surface normals. The relative pixel brightnesses for the different light source positions reveal the correct surface orientations. Finally, the resulting unambiguous surface normal estimates are integrated to recover a depth map. The technique is tested on various objects of different materials. The paper also demonstrates how the depth estimates can be enhanced by applying methods suggested in earlier work.