This work takes part of the development of far-infrared and millimeter astrophysics. We have worked on the data processing and analysis in the fields of the Galactic interstellar medium, through the dust thermal emission, and cosmology through the observation of the cosmic microwave background fluctuations. We have been particularly interested in optimal map-making by inverse linear methods. We have developed a new map-making method for the balloon-borne submillimeter experiment PRONAOS, based on a Wiener inversion matrix, which allows to globally reconstruct the map. The analysis of PRONAOS maps in massive star-forming complexes as Orion and M17 allowed us to discover the large variations of the physical conditions and the dust properties. We showed an anticorrelation between the temperature and the submillimeter spectral index. Our investigations concerning this effect favour causes related to the intrinsic physics of the grains. We have also developed optimal map-making methods for the experiments aiming at measuring the cosmic microwave background fluctuations. The iterative methods that we have developed allow to reconstruct the sky maps very accurately, in spite of the large amount of self-correlated noise present in the timelines. We have also worked on the data processing and analysis for the Archeops balloon-borne experiment.