This thesis discusses the mesospheric retrievals from high spectral resolution infrared limb measurements data as measured by the Michelson Interferometer for Passive Atmospheric Sounding (MIPAS). Altough MIPAS was designed for stratospheric studies, since April 2007, MIPAS measures the entire mesosphere at least every ten days using a special viewing mode. Retrieving geophysical properties at such high altitudes requires careful attention due to signal to noise limitations as well as the intrusion of non Local Thermodynamic Equilibrium (nonLTE) effects. The three main parts of this thesis are: (1) The evaluation of the Oxford MIPAS middle atmosphere dataset (profiles of pressure, temperature, H2O, O3, HNO3, CH4, N2O, NO2 and CO) as retrieved by the iterative algorithm MORSE. Comparisons of daily zonal means, daily zonal sections and comparisons of pairs of profiles against several mesospheric instruments are shown. (2) The description and validation of a new linear (non-iterative) retrieval algorithm. This algorithm exploits the linearity of an optically thin path which allows to use all the available spectral points of the target parameter rather than just subset of the spectra. (3) The validation of the modelled vibrational temperatures used either in the MORSE or in the linear retrieval to estimate the influence of the nonLTE effects upon the measured radiaces. This thesis also briefly investigates the potential of the HIgh Resolution Dynamic Limb Sounder (HIRDLS) instrument to provide valuable mesospheric information, as well as the detection of polar mesospheric clouds using MIPAS radiances.