Catálogo de publicaciones - libros

Symmetry is an important feature in vision. Several detectors or transforms have been proposed. In this paper we concentrate on a measure of symmetry. Given a transform , the kernel of a pattern is defined as the maximal included symmetric sub-set of this pattern. The maximum being taken over all directions, the problem arises to know which center to use. Then the optimal direction triggers the shift problem too. We prove that, in any direction, the optimal axis corresponds to the maximal correlation of a pattern with its flipped version. That leads to an efficient algorithm. As for the measure we compute a modified difference between respective surfaces of a pattern and its kernel. A series of experiments supports actual algorithm validation.

Experiments were carried out to investigate the possibility of training cellular automata to to perform processing. Currently, only binary images are considered, but the space of rule sets is still very large. Various objective functions were considered, and sequential floating forward search used to select good rule sets for a range of tasks, namely: noise filtering, thinning, and convex hulls. Several modifications to the standard CA formulation were made (the B-rule and 2-cycle CAs) which were found to improve performance.

Using a landmark based approach to Procrustes alignment neglects the functional nature of outlines and surfaces. In order to re-introduce this functional nature into the analysis we will consider alignment of shapes with functional representations. First functional Procrustes analysis of curve shapes is treated. Following this we will address the analysis of surface shapes.

The presentation is devoted to the research of mathematical fundamentals for image analysis and recognition procedures. The final goal of this research is automated image mining: a) automated design, test and adaptation of techniques and algorithms for image recognition, estimation and understanding; b) automated selection of techniques and algorithms for image recognition, estimation and understanding; c) automated testing of the raw data quality and suitability for solving the image recognition problem. The main instrument is the Descriptive Approach to Image Analysis, which provides: 1) standardization of image analysis and recognition problems representation; 2) standardization of a descriptive language for image analysis and recognition procedures; 3) means to apply common mathematical apparatus for operations over image analysis and recognition algorithms, and over image models. It is shown also how and where to link theoretical results in the foundations of image analysis with the techniques used to solve application problems.

In this paper we propose a novel method for affine registration of images and point patterns. The method is non-iterative and it directly utilizes the intensity distribution of the images or the spatial distribution of points in the patterns. The method can be used to align images of isolated objects or sets of 2D and 3D points. For Euclidean and similarity transformations the additional contraints can be easily embedded in the algorithm. The main advantage of the proposed method is its efficiency since the computational complexity is only linearly proportional to the number of pixels in the images (or to the number of points in the sets).In the experiments we have compared our method with some other non-feature-based registration methods and investigated its robustness. The experiments show that the proposed method is relatively robust so that it can be applied in practical circumstances.

Demosaicking of the color CCD data is a key to the image quality of digital still and video cameras. Limited by the Nyquist frequency of the color filter array (CFA), color artifacts often accompany high frequency contents in the reconstructed images. This paper presents a general approach of joint spatial-temporal color demosaicking that exploits all three forms of sample correlations: spatial, spectral, and temporal. By motion estimation and statistical data fusion between multiple estimates obtained from adjacent mosaic frames, the new approach can significantly outperform the existing spatial color demosaicking techniques both in objective measure and subjective visual quality.

A template based matching method, adopted to the application of identifying individual proteins of a certain kind in volume images, is presented. Grey-level and gradient magnitude information is combined in the watershed algorithm to extract stable borders. These are used in a subsequent hierarchical matching algorithm. The matching algorithm uses a distance transform to search for local best fits between the edges of a template and edges in the underlying image. It is embedded in a resolution pyramid to decrease the risk of getting stuck in false local minima. This method makes it possible to find proteins attached to other proteins, or proteins appearing as split into parts in the images. It also decreases the amount of human interaction needed for identifying individual proteins of the searched kind. The method is demonstrated on a set of three volume images of the antibody IgG in solution.

Thickness estimation of discrete objects is often a critical step for shape analysis and quantification in medical applications. In this paper, we propose an approach to estimate the thickness (diameter or cross-section area) of discrete tree-like tubular objects in 3D binary images. The estimation is performed by an iterative process involving skeletonization, skeleton simplification, discrete cross-section plane evaluation and finally area estimation. The method is essentially based on discrete geometry concepts (skeleton, discrete planes, and discrete area). It has been validated on phantoms in order to determine its robustness in case of thickness variations along the studied object. The method has also been applied for vessel quantification and computer-aided diagnosis of vascular pathologies in angiographic data, providing promising results.

This paper presents an algorithm for improving the segmentation from a semi-automatic artificial neural network (ANN) hippocampus segmentation of co-registered T1-weigthted and T2-weighted MRI data, in which the semi-automatic part is the selection of a bounding-box. Due to the morphological complexity of the hippocampus and the difficulty of separating from adjacent structures, reproducible segmentation using MR imaging is complicated.

The grey-level thresholding uses a histogram-based method to find robust thresholds. The T1-weighted data is grey-level eroded and dilated to reduce leaking from hippocampal tissue to the surrounding tissues and selecting possible foreground tissue. The method is a 3D approach, it uses 3 × 3 × 3 structure element for the grey-level morphology operations and the algorithms are applied in the three directions, sagittal, axial, and coronal, and the results are then combined together.

In this paper we analyze neural stem/progenitor cells in an time-lapse image sequence. By using information about the previous positions of the cells, we are able to make a better selection of possible cells out of a collection of blob-like objects. As a blob detector we use Laplacian of Gaussian (LoG) filters at multiple scales, and the cell contours of the selected cells are segmented using dynamic programming. After the segmentation process the cells are tracked in the sequence using a combined nearest-neighbor and correlation matching technique. An evaluation of the system show that 95% of the cells were correctly segmented and tracked between consecutive frames.