Susana Mata

Neuronize: a tool for building realistic neuronal cell morphologies

This study presents a tool, Neuronize, for building realistic three-dimensional models of neuronal cells from the morphological information extracted through computer-aided tracing applications. Neuronize consists of a set of methods designed to build 3D neural meshes that approximate the cell membrane at different resolution levels, allowing a balance to be reached between the complexity and the quality of the final model. The main contribution of the present study is the proposal of a novel approach to build a realistic and accurate 3D shape of the soma from the incomplete information stored in the digitally traced neuron, which usually consists of a 2D cell body contour. This technique is based on the deformation of an initial shape driven by the position and thickness of the first order dendrites. The addition of a set of spines along the dendrites completes the model, building a final 3D neuronal cell suitable for its visualization in a wide range of 3D environments.

Using grey-level and shape information for decomposing proteins in 3D images

An image analysis method for decomposing 3D objects using a combination of grey-level and shape is presented. The method consists of two major parts: seeding based on grey-level information and growth from the seeds based on shape information. The growth is performed in two steps in order to prevent seeds located in peripheral or protruding parts of the object from growing into other parts. The method was developed to decompose 3D reconstructions of proteins into their structural subunits. The proteins are imaged with SET (Sidec electron tomography) at a resolution of approximately 2 nm. Decomposition can be a useful tool in the segmentation process to help distinguish between true protein molecules and other objects. It can also be useful for analyzing and visualizing interactions between proteins.

Perceptual evaluation of illumination effects in virtual environments

In this work, different techniques for the generation of shadows and reflections have been compared in terms of the perceived quality by the final user. Results show that, for the analyzed scenarios, users do not present a clear preference for the images generated with the more sophisticated techniques.

Attention-Based mesh simplification using distance transforms

Although widely used for image processing, Distance Transforms have only recently started to be used in computer graphics This paper proposes a new mesh simplification technique based on Distance Transforms that allows taking into account the proximity of a mesh element to the focus of attention for adapting the approximation error which will be tolerated during the simplification process to the relative importance of that mesh element Experimental results show the feasibility of this approach.

Adaptive mesh simplification using distance to focus of attention for multiple points of view

Multiresolution modelling has been widely accepted as a good approach in order to reach a compromise between quality and performance in the rendering of complex geometries. Adaptive simplification techniques that preserve visually outstanding features have also been explored. However, the presence of external factors, such as lights or movement, may also influence visual attention. The goal of this paper is to present a technique based on Distance Transforms that allows to classify the meshs elements according to their proximity to the focus of attention, and to make use of this information for weighting the approximation error which will be tolerated during the mesh simplification process. The approach followed in this work precomputes the correspondence between image pixels and meshs elements for a given set of points of view performing correspondence estimates for other points of view. The results achieved so far show the feasibility of the proposed approach.

Medial grey-level based representation for proteins in volume images

We present an algorithm to extract a medial representation of proteins in volume images. The representation (MGR) takes into account the internal grey-level distribution of the protein and can be extracted without first segmenting the image into object and background. We show how MGR can facilitate the analysis of the structure of the proteins and thereby also classification. Results are shown on two types of protein images.

Combining visual relevance criteria into adaptive mesh simplification techniques

A wide range of simplification techniques for 3D meshes may take advantage of the limitations of the human visual system by concentrating computational resources in those regions where the users attention will focus most often. Some of those visually outstanding regions may be extracted by means of an automatic visual attention model. Additionally, there are some features that are known to have a high impact in the quality perceived by our visual system, as for example the silhouettes. The goal of this paper is to present the feasibility of combining different perceptual criteria into a set of labels that encode the importance of mesh elements according to their distance to relevant regions. These regions may have been independently extracted by applying different criteria in preprocessing time. The combination of the perceptual criteria is parameterizable, allowing to vary their relative importance. Moreover, the possibility of performing this combination in real time allows to adapt the regions of attention to the visualization conditions.

Using Distance Information for Silhouette Preservation in Mesh Simplification Techniques

The goal of this work is to present a multiresolution technique based on Distance Transforms that allows to classify the elements of the mesh according to their proximity to both the internal and the external contours and makes use of this information for weighting the approximation error which will be tolerated during the mesh simplification process. The approach used in this work precomputes silhouettes for a given set of cameras and performs an estimation for any other point of view. The results obtained are evaluated in two ways: visually and using an objective metric that measures the geometrical difference between two polygonal meshes.

Incorporating visual attention into mesh simplification techniques

The goal of this work is to propose a new automatic technique that makes use of the information obtained by means of a visual attention model for guiding the extraction of a simplified 3D model.

Influence of location over several classifiers in 2d and 3d face verification

In this paper two methods for human face recognition and the influence of location mistakes are shown. First one, Principal Components Analysis (PCA), has been one of the most applied methods to perform face verification in 2D. In our experiments three classifiers have been considered to test influence of location errors in face verification using PCA. An initial set of ”correct located faces” has been used for PCA matrix computation and to train all classifiers. An initial test set was built considering a ”correct located faces” set (based on different images than training ones) and then a new test set was obtained by applying a small displacement in both axis (20 pixels) to the initial set. Second method is based on geometrical characteristics constructed with facial and cranial points that come from a 3D representation. Data are acquired by a calibrated stereo system. Classifiers considered for both methods are k-nearest neighbours (KNN), artificial neural networks: radial basis function (RBF) and Support Vector Machine (SVM). Given our data set, results show that SVM is capable to classify correctly in the presence of small location errors. RBF has an acceptable correct rate but the number of false positives is always higher than in the SVM case.