Thermal infrared (TIR) satellite images are of great interest in many remote sensing applications. However, owing to well-known physical reasons and technical limitations, the TIR band has a spatial resolution that is coarser than the other multispectral bands for a given satellite sensor, i.e. visible and shortwave IR. Although the correlation coefficient between the TIR band and the other spectral bands is relatively small, the spatial cross-correlation shows a pattern of joint variability. In this fact, and in the spatial correlation of each spectral band, reside the fundamentals of image fusion by cokriging. A method is showed here for obtaining TIR images with enhanced spatial resolution using the TIR band and one or more spectral bands of the same scene and better spatial resolution. The technique is illustrated with a Landsat 7 ETM+ image. The high resolution TIR image has perfect coherence, if the cokriged image is degraded to its original resolution the degraded image is identical to the original and values of ERGAS (Erreur Relative Globale Adimensionnelle de Synthèse) and average Q (Image Quality Index) equal to 0.220 and 0.986 respectively were obtained.