Many routine medical examinations produce images of patients suffering from various pathologies. With the huge number of medical images, the manual analysis and interpretation became a tedious task. Thus, automatic image segmentation became essential for diagnosis assistance. Segmentation consists in dividing the image into homogeneous and significant regions. We focus on hidden Markov random fields referred to as HMRF to model the problem of segmentation. This modelisation leads to a classical function minimisation problem. Broyden–Fletcher–Goldfarb–Shanno algorithm referred to as BFGS is one of the most powerful methods to solve unconstrained optimisation problem. In this paper, we investigate the combination of HMRF and BFGS algorithm to perform the segmentation operation. The proposed method shows very good segmentation results comparing with well-known approaches. The tests are conducted on brain magnetic resonance image databases (BrainWeb and IBSR) largely used to objectively confront the results obtained. The well-known Dice coefficient (DC) was used as similarity metric. The experimental results show that, in many cases, our proposed method approaches the perfect segmentation with a Dice Coefficient above .9. Moreover, it generally outperforms other methods in the tests conducted.

The goal of image segmentation is to simplify the representation of an image to items meaningful and easier to analyze. Medical image segmentation is one of the fundamental problems in image processing field. It aims to provide a crucial decision support to physicians. There is no one way to perform the segmentation. There are several methods based on HMRF. Hidden Markov Random Fields (HMRF) constitute an elegant way to model the problem of segmentation. This modelling leads to the minimization of an energy function. In this paper we investigate direct search methods that are Nelder-Mead and Torczon methods to solve this optimization problem. The quality of segmentation is evaluated on grounds truths images using the Kappa index called also Dice Coefficient (DC). The results show the supremacy of the methods used compared to others methods.

International audience

This paper presents a new solver for systems of nonlinear equations. Such systems occur in Geometric Constraint Solving, e.g., when dimensioning parts in CAD-CAM, or when computing the topology of sets defined by nonlinear inequalities. The paper does not consider the problem of decomposing the system and assembling solutions of subsystems. It focuses on the numerical resolution of well-constrained systems. Instead of computing an exponential number of coefficients in the tensorial Bernstein basis, we resort to linear programming for computing range bounds of system equations or domain reductions of system variables. Linear programming is performed on a so called Bernstein polytope: though, it has an exponential number of vertices (each vertex corresponds to a Bernstein polynomial in the tensorial Bernstein basis), its number of hyperplanes is polynomial: O(n2) for a system in n unknowns and equations, and total degree at most two. An advantage of our solver is that it can be extended to non-algebraic equations. In this paper, we present the Bernstein and LP polytope construction, and how to cope with floating point inaccuracy so that a standard LP code can be used. The solver has been implemented with a primal-dual simplex LP code, and some implementation variants have been analyzed. Furthermore, we show geometric-constraint-solving applications, as well as numerical intersection and distance computation examples.

Classically, geometric constraint solvers use graph-based methods to analyze systems of geometric constraints. These methods have intrinsic limitations, which the witness method overcomes. This paper details the computation of a basis of the vector space of the free infinitesimal motions of a typical witness, and explains how to use this basis to interrogate the witness for detecting all dependencies between constraints: structural dependencies already detectable by graph-based methods, and also non-structural dependencies, due to known or unknown geometric theorems, which are undetectable with graph-based methods. The paper also discusses how to decide about the rigidity of a witness.

This paper presents some important issues and potential research tracks for Geometric Constraint Solving: the use of the simplicial Bernstein base to reduce the wrapping effect in interval methods, the computation of the dimension of the solution set with methods used to measure the dimension of fractals, the pitfalls of graph based decomposition methods, the alternative provided by linear algebra, the witness configuration method, the use of randomized provers to detect dependences between constraints, the study of incidence constraints, the search for intrinsic (coordinate-free) formulations and the need for formal specifications.

Geometric constraint solving is a key issue in CAD/CAM. Since Owen's seminal paper, solvers typically use graph based decomposition methods. However, these methods become difficult to implement in 3D and are misled by geometric theorems. We extend the Numerical Probabilistic Method (NPM), well known in rigidity theory, to more general kinds of constraints and show that NPM can also decompose a system into rigid subsystems. Classical NPM studies the structure of the Jacobian at a random (or generic) configuration. The variant we are proposing does not consider a random configuration, but a configuration similar to the unknown one. Similar means the configuration fulfills the same set of incidence constraints, such as collinearities and coplanarities. Jurzak's prover is used to find a similar configuration.

Finite-precision leads to many problems in geometric methods from CAD or Computational Geometry. Until now, using exact rational arithmetic was a simple, yet much too slow, solution to be of any practical use in real-scale applications. A recent optimization-the lazy rational arithmetic-seems promising: It defers exact computations until they become either unnecessary (in most cases) or unavoidable; in such a context, only indispensable computations are performed exactly, that is, those without which any given decision cannot be reached safely using only floating-point arithmetic. This paper takes stock of the lazy arithmetic paradigm: principles, functionalities and limits, speed, possible variants and extensions, difficulties, problems solved or left unresolved.

Numerous methods have been proposed in order to solve geometric constraints, all of them having their own advantages and drawbacks. In this article, we propose an enhancement to the classical numerical methods, which, up to now, are the only ones that apply to the general case.

Symbolic perturbation by infinitely small values removes degeneracies in geometric algorithms and enables programmers to handle only generic cases: there are a few such cases, whereas there are an overwhelming number of degenerate cases. Current perturbation schemes have limitations. To overcome them, the paper proposes to use an /spl epsiv/-arithmetic, i.e. to represent in an explicit way infinitely small numbers and to define arithmetic operations (+,-,*,/,>

Systems based on exact arithmetic are very slow. In practical situations, very few computations need be performed exactly as approximating the results is very often sufficient. Unfortunately, it is impossible to know at the time when the computation is called for whether an exact evaluation will be necessary or not. The arithmetic library presented here achieves laziness by postponing any exact computation until it is proved to be indispensable. This yields very substantial gains in performance while allowing exact decisions. The lazy arithmetic techniques are presented in the context of rational computations, using the field of computational geometry as a background.>