Catálogo de publicaciones - libros

This paper describes an experimental study about a robust contour feature () for using in action recognition based on continuous hidden Markov models (HMM). We ran different experimental setting using the KTH’s database of actions. The image contours are extracted using a standard algorithm. The feature vector is build from of histogram of a set ofnon-overlapping regions in the image. We show that the combined use of HMM and this feature gives equivalent o better results, in term of action detection, that current approaches in the literature.

This paper presents a new approach to the problem of simultaneous location and segmentation of object in images. The main emphasis is done on the information provided by the contour fragments present in the image. Clusters of contour fragments are created in order to represent the labels defining the different parts of the object. An unordered probabilistic graph is used to model the objects, where a greedy approach (using dynamic programming) is used to fit the graph model to the labels.

A new method is proposed for improving microarray spot segmentation for gene quantification. The method introduces a novel combination of three image processing stages, applied locally to each spot image: i/ Fuzzy C-Means unsupervised clustering, for automatic spot background noise estimation, ii/ power spectrum deconvolution filter design, employing background noise information, for spot image restoration, iii/ Gradient-Vector-Flow (GVF-Snake), for spot boundary delineation. Microarray images used in this study comprised a publicly available dataset obtained from the database of the MicroArray Genome Imaging & Clustering Tool website. The proposed method performed better than the GVF-Snake algorithm (Kullback-Liebler metric: 0.0305 bits against 0.0194 bits) and the SPOT commercial software (pairwise mean absolute error between replicates: 0.234 against 0.303). Application of efficient adaptive spot-image restoration on cDNA microarray images improves spot segmentation and subsequent gene quantification.

On–board pedestrian detection is a key task in advanced driver assistance systems. It involves dealing with aspect–changing objects in cluttered environments, and working in a wide range of distances, and often relies on a classification step that labels image regions of interest as pedestrians or non–pedestrians. The performance of this classifier is a crucial issue since it represents the most important part of the detection system, thus building a good classifier in terms of false alarms, missdetection rate and processing time is decisive. In this paper, a pedestrian classifier based on Haar wavelets and edge orientation histograms (HW+EOH) with AdaBoost is compared with the current state–of–the–art best human–based classifier: support vector machines using histograms of oriented gradients (HOG). The results show that HW+EOH classifier achieves comparable false alarms/missdetections tradeoffs but at much lower processing time than HOG.

This paper describes how face recognition can be effected using 3D shape information extracted from single 2D image views. We characterise the shape of the field of facial normals using a statistical model based on principal geodesic analysis. The model can be fitted to 2D brightness images of faces to recover a vector of shape parameters. Since it captures variations in a field of surface normals, the dimensionality of the shape vector is twice the number of image pixels. We investigate how to perform face recognition using the output of PGA by applying a number of dimensionality reduction techniques including principal components analysis, locally linear embedding, locality preserving projection and Isomap.

Performance in most pattern classifiers is improved when redundant or irrelevant features are removed, however, this is mainly achieved by high demanding computational methods or successive classifiers construction. This paper shows how Associative Memories can be used to get a mask value which represents a subset of features that clearly identifies irrelevant or redundant information for classification purposes, therefore, classification accuracy is improved while significant computational costs in the learning phase are reduced. An optimal subset of features allows register size optimization, which contributes not only to significant power savings but to a smaller amount of synthesized logic, furthermore, improved hardware architectures are achieved due to functional units size reduction, as a result, it is possible to implement parallel and cascade schemes for pattern classifiers on the same ASIC.

The main goal of this work is to study the performance of CARFIR (Automatic Construction of Rules in Fuzzy Inductive Reasoning) methodology for the modelling and prediction of the human central nervous system (CNS). The CNS controls the hemodynamical system by generating the regulating signals for the blood vessels and the heart.CARFIR is able to automatically construct fuzzy rules starting from a set of pattern rules obtained by FIR. The methodology preserves as much as possible the knowledge of the pattern rules in a compact fuzzy rule base. The prediction results obtained by the fuzzy prediction process of CARFIR methodology are compared with those of other inductive methodologies, i.e. FIR, NARMAX and neural networks.

In this work, a new method for detecting copies of a query video in a videos database is proposed. It includes a new clustering technique that groups frames with similar visual content, maintaining their temporal order. Applying this technique, a keyframe is extracted for each cluster of the query video. Keyframe choice is carried out by selecting the frame in the cluster with maximum similarity to the rest of frames in the cluster. Then, keyframes are compared to target videos frames in order to extract similarity regions in the target video. Relaxed temporal constraints are subsequently applied to the calculated regions in order to identify the copy sequence. The reliability and performance of the method has been tested by using several videos from the MPEG-7 Content Set, encoded with different frame sizes, bit rates and frame rates. Results show that our method obtains a significant improvement with respect to previous approaches in both achieved precision and computation time.

Object recognition requires that templates with canonical views are stored in memory. Such templates must somehow be normalised. In this paper we present a novel method for obtaining 2D translation, rotation and size invariance. Cortical simple, complex and end-stopped cells provide multi-scale maps of lines, edges and keypoints. These maps are combined such that objects are characterised. Dynamic routing in neighbouring neural layers allows feature maps of input objects and stored templates to converge. We illustrate the construction of group templates and the invariance method for object categorisation and recognition in the context of a cortical architecture, which can be applied in computer vision.

One area of pattern recognition that is receiving a lot of attention recently is handwritten text recognition. Traditionally, handwritten text recognition systems have been modelled by means of HMM models and n-gram language models. The problem that n-grams present is that they are not able to capture long-term constraints of the sentences. Stochastic context-free grammars (SCFG) can be used to overcome this limitation by rescoring a n-best list generated with the HMM-based recognizer. Howerver, SCFG are known to have problems in the estimation of comlpex real tasks. In this work we propose the use of a combination of n-grams and category-based SCFG together with a word distribution into categories. The category-based approach is thought to simplify the SCFG inference process, while at the same time preserving the description power of the model. The results on the IAM-Database show that this combined scheme outperforms the classical scheme.