Spatial resolution is one of the most critical measurement parameters in infrared microspectroscopy. Due to the distinct levels of morphologic heterogeneity in cells and tissues the spatial resolution in a given IR imaging setup strongly affects the character of the infrared spectral patterns obtained from the biomedical samples. This is particularly important when spectral data bases of reference microspectra from defined tissue structures are collected. In this paper we have also pointed out that the concept of spatial resolution in IR imaging is inseparable from the contrast. Based on infrared microspectroscopic transmittance data acquired from an USAF 1951 resolution target we have demonstrated how the spatial resolution can be determined experimentally and some numbers for the spatial resolution of popular IR imaging systems are provided. Finally, we have presented a new computational procedure which is suitable to improve the spatial resolution in IR imaging. A theoretical model of 3D-Fourier self-deconvolution (FSD) is given and advantages or pitfalls of this method are discussed. Based on synchrotron IR microspectroscopic data we have furthermore demonstrated that the technique of 3D-FSD can be successfully applied to increase the spatial resolution in a real IR imaging setup.