Catálogo de publicaciones - libros

A new method for the reduction of the number of colors in a digital image is proposed. The new method is based on the development of a new neural network classifier that combines the advantages of the Growing Neural Gas (GNG) and the Kohonen Self-Organized Feature Map (SOFM) neural networks. We call the new neural network: Self-Growing and Self-Organized Neural Gas (SGONG). Its main advantage is that it defines the number of the created neurons and their topology in an automatic way. Besides, a new method is proposed for the Estimation of the Most Important of already created Classes (EMIC). The combination of SGONG and EMIC in color images results in retaining the isolated and significant colors with the minimum number of color classes. The above techniques are able to be fed by both color and spatial features. For this reason a similarity function is used for vector comparison. To speed up the entire algorithm and to reduce memory requirements, a fractal scanning sub-sampling technique is used. The method is applicable to any type of color images and it can accommodate any type of color space.

We introduce a method to reduce oversegmentation in watershed partitioned images, that is based on the use of a multiresolution representation of the input image. The underlying idea is that the most significant components perceived in the highest resolution image will remain identifiable also at lower resolution. Thus, starting from the image at the highest resolution, we first obtain a multiresolution representation by building a resolution pyramid. Then, we identify the seeds for watershed segmentation on the lower resolution pyramid levels and suitably use them to identify the significant seeds in the highest resolution image. This is finally partitioned by watershed segmentation, providing a satisfactory result. Since different lower resolution levels can be used to identify the seeds, we obtain alternative segmentations of the highest resolution image, so that the user can select the preferred level of detail.

This paper presents a novel approach for image retrieval from digital collections. Specifically, we describe IRONS (Image Retrieval with Ontological Descriptions of Shapes), a system based on the application of several novel algorithms that combine low-level image analysis techniques with automatic shape extraction and indexing. In order to speed up preprocessing, we have proposed and implemented the convex regions algorithm and discrete curve evolution approach. The image indexing module of IRONS is addressed using two proposed algorithms: the tangent space and the two-segment turning function for shapes representation invariant to rotation and scale. Another goal of the proposed method is the integration of user-oriented descriptions, which leads to more complete retrieval by accelerating the convergence to the expected result. For the definition of image semantics, ontology annotation of sub-regions has been used.

Most of the document binarization techniques have many parameters that can initially be specified. Usually, subjective document binarization evaluation, employs human observes for the estimation of the best parameter values of the techniques. Thus, the selection of the best values for these parameters is crucial for the final binarization result. However, there is not any set of parameters that guarantees the best binarization result for all document images. It is important, the estimation of the best values to be adaptive for each one of the processing images. This paper proposes a new method which permits the estimation of the best parameter values for each one of the document binarization techniques and also the estimation of the best document binarization result of all techniques. In this way, document binarization techniques can be compared and evaluated using, for each one of them, the best parameter values for every document image.

In this paper we present a new method for comparing histograms. Its main advantage is that it takes less time than previous methods.

The present distances between histograms are defined on a structure called signature, which is a lossless representation of histograms. Moreover, the type of the elements of the sets that the histograms represent are ordinal, nominal and modulo.

We show that the computational cost of these distances is (′) for the ordinal and nominal types and () for the modulo one, where ′ is the number of non-empty bins of the histograms. In the literature, the computational cost of the algorithms presented depends on the number of bins in the histograms. In most applications, the histograms are sparse, so considering only the non-empty bins dramatically reduces the time needed for comparison.

The distances we present in this paper are experimentally validated on image retrieval and the positioning of mobile robots through image recognition.

Median associative memories (MEDMEMs) first described in [1] have proven to be efficient tools for the reconstruction of patterns corrupted with mixed noise. First formal conditions under which these tools are able to reconstruct patterns either from the fundamental set of patterns and from distorted versions of them were given in [1]. In this paper, new more accurate conditions are provided that assure perfect reconstruction. Numerical and real examples are also given.

Automatic Indexing of Broadcast News is a developing research area of great recent interest [1]. This paper describes the development steps for designing an automatic index system of broadcast news for both Basque and Spanish. This application requires of appropriate Language Resources to design all the components of the system. Nowadays, large and well-defined resources can be found in most widely used languages, but there is a lot of work to do with respect to minority languages. Even if Spanish has much more resources than Basque, this work has parallel efforts for both languages. These two languages have been chosen because they are evenly official in the Basque Autonomous Community and they are used in many mass media of the Community including the Basque Public Radio and Television EITB [2].

We address the problem of pose and illumination invariance in face recognition and propose to use explicit 3D model and variants of existing algorithms for both pose [Fit01, MSCA04] and illumination normalization [ZS04] prior to applying 2D face recognition algorithm. However, contrary to prior work we will use person specific, rather than general 3D face models. The proposed solution is realistic as for many applications the additional cost of acquiring 3D face images during enrolment of the subjects is acceptable. 3D sensing is not required during normal operation of the face recognition system. The proposed methodology achieves illumination invariance by estimating the illumination sources using the 3D face model. By-product of this process is the recovery of the face skin albedo which can be used as a photometrically normalised face image. Standard face recognition techniques can then be applied to such illumination corrected images.

Automatic semantic annotation of video streams allows to extract significant clips for archiving and retrieval of video content. In this paper, we present a system that performs automatic annotation of soccer videos, detecting principal highlights, and recognizing identity of players. Highlight detection is carried out by means of finite state machines that encode domain knowledge, while player identification is based on face detection, and on the analysis of contextual information such as jersey’s numbers and superimposed text captions. Results obtained on actual soccer videos shows overall highlight detection rates of about 90%. Lower, but still promising, accuracy is achieved on the very difficult player identification task.

In this paper the authors use the framework of conformal geometric algebra for the treatment of robot vision tasks. In this mathematical system we calculated projective invariants using omnidirectional vision for object recognition. We show the power of the mathematical system for handling differential kinematics in visual guided tracking.