Sensitivity to mechanical vibrations of Fourier Transform Spectrometers (FTS) is a well-known phenomenon. It is especially critical for FTS devoted to atmospheric studies (like the Planetary Fourier Spectrometer (PFS) onboard Mars Express 2003), as absorption bands for the gases of low concentration are comparable with the generated instrument spectral noise. The adopted techniques for the vibration sensitivity reduction suffer of limitations in practical implementation, leaving residual modulations of the interferogram and the so-called ghosts in the spectra. Moreover as it is often impossible to measure the vibrations during the FTS measurement, the position and magnitude of these ghosts cannot be evaluated. Up to now the adopted ghost reduction techniques are mostly based on the averaging of spectra, because the disturbance phase is randomly distributed. This paper presents an innovative data treatment technique which allows single spectrum correction from distortions of unknown nature. Such a technique would increase the spatial resolution of the mapping process and becomes crucial when the desired information is linked to a particular mapping area associated to an individual spectrum. The full study consists in the explicit analysis of the ghost formation and the post-processing algorithm based on the semiblind deconvolution method - an iterative numerical algorithm of the series of consecutive deconvolutions. The technique was tested on the data from the PFS and the algorithm proved to be consistent according to the selected efficiency criteria (coming from the available general information about the signal spectral shape). © (2013) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.