Catálogo de publicaciones - libros

Recently, an energy-based unified framework for image denoising was proposed by Mrázek et al. [10], from which existing nonlinear filters such as M-smoothers, bilateral filtering, diffusion filtering and regularisation approaches, are obtained as special cases. Such a model offers several degrees of freedom (DOF) for tuning a desired filter. In this paper, we explore the generality of this filtering framework in combining nonlocal tonal and spatial kernels. We show that Bayesian analysis provides suitable foundations for restricting the parametric space in a noise-dependent way. We also point out the relations among the distinct DOF in order to guide the selection of a combined model, which itself leads to hybrid filters with better performance than the previously mentioned special cases. Moreover, we show that the existing trade-off between the parameters controlling similarity and smoothness leads to similar results under different settings.

Automated brain MR image segmentation is a challenging problem and received significant attention lately. Various techniques have been proposed, several improvements have been made to the standard fuzzy c-means (FCM) algorithm, in order to reduce its sensitivity to Gaussian, impulse, and intensity non-uniformity noises. In this paper we present a modified FCM algorithm, which aims at accurate segmentation in case of mixed noises, and performs at a high processing speed. As a first step, a scalar feature value is extracted from the neighborhood of each pixel, using a filtering technique that deals with both spatial and gray level distances. These features are clustered afterwards using the histogram-based approach of the enhanced FCM algorithm. The experiments using 2-D synthetic phantoms and real MR images show, that the proposed method provides better results compared to other reported FCM-based techniques. The produced segmentation and fuzzy membership values can serve as excellent support for level set based cortical surface reconstruction techniques.

Mottling is one of the most significant defects in modern offset printing using coated papers. Mottling can be defined as undesired unevenness in perceived print density. Previous research in the field considered only gray scale prints. In our work, we extend current methodology to color prints. Our goal was to study the characteristics of the human visual system, perform psychometric experiments and develop methods which can be used at industrial level applications. We developed a method for color prints and extensively tested it with a number of experts and laymen. Suggested approach based on pattern-color perception separability proved to correlate with the human evaluation well.

Cell cultures as well as cells in tissue always display a certain degree of variability, and measurements based on cell averages will miss important information contained in a heterogeneous population. This paper presents automated methods for image based measurements of mitochondiral DNA (mtDNA) mutations in individual cells. The mitochondria are present in the cell’s cytoplasm, and each cytoplasm has to be delineated. Three different methods for segmentation of cytoplasms are compared and it is shown that automated cytoplasmic delineation can be performed 30 times faster than manual delineation, with an accuracy as high as 87%. The final image based measurements of mitochondrial mutation load are also compared to, and show high agreement with, measurements made using biochemical techniques.

Moving objects detection is a crucial step for video surveillance systems. The segmentation performed by motion detection algorithms is often noisy, which makes it hard to distinguish between relevant motion and noise motion. This article describes a new approach to make such a distinction using principal component analysis (PCA), a technique not commonly used in this domain. We consider a ten-frame subsequence, where each frame is associated with one dimension of the feature space, and we apply PCA to map data in a lower-dimensional space where points picturing coherent motion are close to each other. Frames are then split into blocks that we project in this new space. Inertia ellipsoids of the projected blocks allow us to qualify the motion occurring within the blocks. The results obtained are encouraging since we get very few false positives and a satisfying number of connected components in comparison to other tested algorithms.

We describe and evaluate a method to robustly detect the page frame in document images, locating the actual page contents area and removing textual and non-textual noise along the page borders. We use a geometric matching algorithm to find the optimal page frame, which has the advantages of not assuming the existence of whitespace between noisy borders and actual page contents, and of giving a practical solution to the page frame detection problem without the need for parameter tuning. We define suitable performance measures and evaluate the algorithm on the UW-III database. The results show that the error rates are below 4% for each of the performance measures used. In addition, we demonstrate that the use of page frame detection reduces the optical character recognition (OCR) error rate by removing textual noise. Experiments using a commercial OCR system show that the error rate due to elements outside the page frame is reduced from 4.3% to 1.7% on the UW-III dataset.

In this article we present a way of representing pairs of orientations in the plane. This is an extension of the familiar way of representing single orientations in the plane. Using this framework, pairs of lines can be added, scaled and averaged over in a sense which is to be described. In particular, single lines can be incorporated and handled simultaneously.

Chamfer matching is an edge based matching technique that has been used in many applications. The matching process is to minimize the distance between transformed model edges and image edges. This distance is usually computed at the pixel resolution using a distance transform, thus reducing accuracy of the matching. In this paper, an improved approach for accurate chamfer matching is presented that uses interpolation in the distance calculation for subpixel distance evaluation. Also, instead of estimating the optimal position in subpixel using a neighborhood of the pixel position with the minimum distance, for more accurate matching, we use the Powell’s optimization to find the distance minimum through actual distance evaluations in subpixel. Experimental results are presented to show the validity of our approach.

In this paper, a segmentation algorithm is proposed which extracts regions depicting fibroglandular tissue in a mammogram. There has been an increasing need for such algorithms due to several reasons, the majority of which are related to the development of techniques for Computer Aided Diagnosis of breast cancer from mammograms. The proposed algorithm consists of a major phase and a post-processing phase. The purpose of the major phase is to calculate the threshold value that yields a segmentation of glandular tissue which is achievable by thresholding. The method by which we calculate this threshold value is based on the principle of minimizing the cross-entropy between two images. The resulting segmentation is then post-processed to remove artifacts such as noise and other unwanted regions. The algorithm has been implemented and evaluated with promising results. In particular, its performance seems to match that of medical professionals specialized in mammography.

This work presents an automatic scenario recognition system for video sequence interpretation. The recognition algorithm is based on a Bayesian Networks approach. The model of scenario contains two main layers. The first one enables to highlight atemporal events from the observed visual features. The second layer is focused on the temporal reasoning stage. The temporal layer integrates an event based approach in the framework of the Bayesian Networks. The temporal Bayesian network tracks lifespan of relevant events highlighted from the first layer. Then it estimates qualitative and quantitative relations between temporal events helpful for the recognition task. The global recognition algorithm is illustrated over real indoor images sequences for an abandoned baggage scenario.