[en] This thesis work is dedicated to the problem of the search and the characterisation of dark matter. From an observational point of view, on different length scale, from galactic to the entire Universe, it exists a disagreement between the dynamical and luminous estimation of the mass of astrophysical objects (as galaxies and cluster of galaxies). This is a sort of missing mass problem. This makes necessary, in the framework of the standard cosmological model, the introduction of a dark component of mass, dark meaning that it doesn't emit electromagnetic radiation and whose presence can be detected through its gravitational effects. Even if both theory and observation agree on the need of such a component, a still unanswered question is the nature of such mass component. We are going to address the problem of the dark matter component in galactic haloes, where the observational evidences in this sense (rotation curves) are very robust. To this end we use an original set of data (the MDM data) collected with the aim of studying the dark component in form of MACHOs (Massive Astrophysical Compact Halo Objects) in the halo of our and of the nearby galaxy of Andromeda (M31). This search is based on the gravitational microlensing effect, the deflection of light due to the presence of a massive body (the MACHO) along the line of sight between the observer and a luminous source. This effects shows as a temporal luminosity variation of the source. From this analysis it is possible to estimate the distribution of the mass in form of MACHO in the halo. This search has been carried out using the pixel lensing technique (proposed and developed by the AGAPE collaboration), that allows one to detect luminosity variations of unresolved sources. In the first chapter we recall some basics points about the cosmological model and on the problem of dark matter. We the review the microlensing effect and the pixel lensing technique. Chapter two is devoted to the description of the experimental setup, and to different points linked with the image analysis (calibration, composition). In chapter three we deal with the techniques used to analyse the signal with respect to the problem of the detection of microlensing event. The background noise being given here by variable stars, in particular we study how we can characterize the achromaticity properties of the signal that interest us. We then discuss the results from two simulations: the first one aimed at the study of the selection criteria we use to detect interesting variations, the second a Monte Carlo simulation of the experience. In chapter 4 we consider various aspects of the analysis. Dealing with the chromatic effects, we study some kind of variable sources. We carry out an extension of known microlensing candidates detected by other collaborations. Eventually, we come to the discussion of the results of our selection, where we get five light curves compatible with a microlensing signal. Looking at the simulation, we then draw our conclusion on the problem. Our analysis tends to confirm that only a small fraction of the galactic haloes is in form of MACHO, and in particular to exclude as a major component substellar objects. (author)