Image Processingn

Work has been carried out on the theoretical and practical development of algorithms applied to biomedicine, which is a very fertile area. In particular it is worth noting the work to analyze images of the retina, improving and tomographic neuroimaging methods. These areas will be further developed at the same time promoting related areas as filtering, segmentation and visualization of 3D images resulting from both models as retinal tomography. It also plans to extend the application areas to fields that produce images that share similar characteristics to those of biomedicine.

Project listing:

Mathematical and Computational Methods for Electronic Miscroscopy of Specimens

Mathematical and Computational Methods for Electronic Miscroscopy of Specimens

The aim is to contribute to the knowledge of the functioning of biological processes, at the subcellular and molecular level, with the development and application of mathematical and engineering methods. As a result, it is intended to produce better images of biological structures that lead to a better understanding of such structures.
Image Mosaic Generation from Eye Background

Image Mosaic Generation from Eye Background

We study the generation of image mosaics of human fundus, in order to obtain a panoramic view of the blood vessels. Graph-based techniques are used to obtain graph showing the correlation between images and techniques for mapping spatial spline therebetween. If successful in this approach in the plane, the respective extension will be in 3D.
Fundus Imaging Processing for the Analysis of Retinopathy in Premature Infants

Fundus Imaging Processing for the Analysis of Retinopathy in Premature Infants (ROP)

We study the segmentation and measurement of blood vessels in premature infants. The current system that segments and measures blood vessels (RISA) was developed for adult images, but the images of infants have a lower resolution and their signal-to-noise ratio is very low, so the RISA system does not have a good performance. To solve this problem, we are developing more robust segmentation systems that are based on fuzzy logic.
Medical Imaging Detection Segmentation and Measurement of the Morphology of Bifurcating Structures

Medical Imaging. Detection, Segmentation, and Measurement of the Morphology of Bifurcating Structures

This work is divided into three main areas: 1) segmentation, 2) measurement of tree structures morphology and 3) 3D reconstruction of these structures using computer vision techniques. The tree segmentation in 2D structures (lines) are based on feature extraction methods second order at different scales using Gaussian filters and region growing techniques. The 3D segmentation methods (tubulars) are based on the same methods but estereopar images and/or multivistas with which the projections are made and the 3D reconstruction. The morphology measurement is performed on binary images, segmentation result described above, and consists of measuring the geometry of trees (usually binary) for data with information from long, diameters, branching angles and relationship between parent and child along the tree, and then derive a number of other geometric measures useful for the description of such structures and the topology of these trees, both 2D and 3D. Applications range by the time two main areas: 1) study photographs of blood vessels in fundus clinics, in order to develop systems for early diagnosis of diseases such as hypertension, diabetes or retinopathy, 2) extraction and measurement of neural trees in confocal microscopy images of neurons, for measurement and counting of synapses, dendrites and axons, and tree topology, in order to support basic research in neuroscience in areas such as memory and learning as well as cell death.


Grupo Visp