An a posteriori tracing method for broken neurite reconstruction is proposed. It uses local line structure cues from three vector fields; diffusion tensor vector field, Hessian eigenvector field, and diffused gradient vector field. The vector fields are built to give directionalities of line structures with gaps in different configurations. The method uses an iterative search and extend approach to reconstruct possible links between broken neurite segments which for various reasons existing tracing methods fail to extract. Iterative nature of the method enables us to extend traced neurites in directions aligned with underlying line structures. Subsequently, the method connects traced neurites across gaps in presence of other traced neurite segments or disposes extensions otherwise.

In the past few years, many change detection methods have been proposed in the GIS community to support the update of existing building databases. If most of these methods use images to settle the problem, only a few of them analyse the optimal resolution to use for input images. The main goal of this paper - that focuses on the use of satellite images - is to provide possible answers to this question. For that purpose, a change detection method is tested on several test areas, for which input images are available at different resolutions (50 cm and 70 cm). The outcomes of the method are then evaluated and analysed in order to highlight the impact of the spatial resolution of input images on the system's efficiency.

In this paper, we present an automatic method for deriving Digital Terrain Models (DTM) from Digital Surface Models (DSM). The DTM generation is formalized here as the minimization of a given energy, defined by a data term and a regularization term. The main contribution of this paper lies in the specific module, introduced in the data term in order to filter outliers (non ground points present in the original DSM) so that the final terrain surface best fits true ground points (inliers). The method is here introduced and tested on a vegetated and hilly area. A comparison to ground truth is also performed and shows the geometric accuracy and the robustness of the approach.

We propose two algorithms to match edges in a geometrically-imprecise graph to geometrically-precise strong boundaries in an image, where the graph is meant to give an a priori partition of the image into objects. This can be used to partition an image into objects described by imprecise external data, and thus to simplify the segmentation problem. We apply them to the problem of registering cadastre data to georeferenced aerial images, thus correcting the lack of geometrical detail of the cadastre data, and the fact that cadastre data gives information of a different nature than that found in images (fiscal information as opposed to actual land use).

This research work is a part of a global project related to urban scene modeling from high resolution satellite images with focus on building production. The input data consist of a panchromatic stereo pair of satellite images, with a submetric resolution of 45-70 cm and a low base to height ratio B/H [0.05 - 0.2]. Data are provided by CNES. Since a detailed extraction and description of building rooves is complex in a satellital context, we propose to describe the scene by means of a 3D-surface which provides either raster or vector information using different level description. From extracted primitives and raw correlation information, an hybrid 3D surface will be obtained by an energy minimization process via graph cuts. The main contribution of our approach is the use of 3D-primitives such as 3D-segments and 3D-facets as well as the introduction of information from external database to guide the optimization process. The obtained product is an hybrid DEM (digital elevation model) which provides the highest level of reliable primitives for each scene region.

Algorithms are presented which allow us to perform morphological transformations on unorganised sets of points represented by their Delaunay triangulation. The results show that these algorithms could behave as morphological operators such as erosion, dilatation, and opening do. They actually act as "shape filters". These algorithms are applied to the problem of scene reconstruction by stereoscopy in which objects are represented by unstructured clouds of 3D points.

One of the major goals of computer vision is the research and the development of flexible methods for shape description. A large group of shape description techniques is given by heuristic approaches, which yield acceptable results in the description of simple shapes and regions. In this case, objects are represented by a planar graph with nodes symbolizing subregions from region decomposition, and region shape is then described by the graph properties. In the paper, the angular bisector network (ABN), a descriptor of polygonal shape, is used to automatically detect intersections between neurites of cell structures. Some properties of the ABN, such as linear algebraic complexity, easy extraction of characteristic points, etc., are very useful and experimental results are promising.

We introduce the new concept of "veinerization", which produces a graph that contains all the "topological" information needed to derive a wide variety of skeletons. Theoretically, the main contribution is to provide a homogeneous framework for integration of the major concepts described in other related works on digital skeletonization. In practice, the new aspect of this approach is to provide the user with different criteria for selecting the most suitable skeleton for a given application, e.g., the user can select a suitable threshold for obtaining the desirable balance between " having a skeleton without noisy prunes" and "having a skeleton that reflects the initial shape". This algorithm has been tested on numerous kinds of patterns, including pathological ones like fractal sets well-known for the complexity of their shapes.

The airborne flash SAR is a versatile device to produce high resolution radar images. In order to produce ortho-images, the distortions due to the complex real trajectory, system errors, and terrain elevation must be rectified. We have conceived (and previously presented) a rectification method specific to the flash radar. To estimate the quality of the rectification and the limit of our method (and of the geometric model) we have used a very accurate trajectory (DGPS) to produce undistorted images. The comparison of these images with the ones obtained by our method shows that our method works well and that the main limit to the rectification is the accuracy of the cartographic data used.

We present a complete method for character string extraction that is specifically adapted to a cartographic context (orientation,...). We emphasize a high level reconstruction process that: (1) resolves the ambiguities remaining from pattern analysis; (2) structures the characters into strings. The knowledge used by this process is the coherence of strings to be constructed (e.g. in orientation, scale, fonts...). We formalize the problem in the most general way and then admit two simplifying hypotheses that transform the problem into a graph optimization problem. The method implemented for solving this problem uses different techniques such as: graph theory, dynamic programming, heuristic, combinatorial exploration.... The results seem sufficiently good for an industrial application.

The cartographic use of radar images and multisensor images fusion requires rectified images. The geometric errors of a radar image are induced by flight and system errors, as well as terrain elevation. The flash radar is a fast processing radar that acquires an image strip by strip, with a position dependent resolution. The ortho-rectification process must then deal with two different problems: correct each strip, and merge them. Due to the small number of available ground control points, and overlapping areas between strips, these two steps must be processed together. The paper presents two types of specific methods to correct flash radar images: interpolative methods, based on polynomial least square fitting, and parametric methods, based on nonlinear least square adjustment of a mathematical model of the flash radar geometric relations. A comparison is made between the first results obtained with these methods.>

The authors describe a high level method for reconstruction of a road network graph extracted from scanned maps. The corrupted graph that results from low-level treatments contains all the information needed for a correct reconstruction if one takes into account the a priori knowledge. A list of rules that contains this knowledge, is given. Formally, the problem is to construct the most likely graph considering a priori knowledge and the corrupted graph with which one started. An algorithmic approach is presented. The authors describe the different algorithmic steps used; each step uses a specific subset of the rules considered. As much as possible, to proceed by global optimization inside each step. The results obtained are good enough for use in GISs (geographic information systems) except for cities.>