A new approach is proposed for the anisotropic diffusion of multispectral imagery. This new technique is called multi-spectral anisotropic diffusion (MSAD) and provides a decrease in computational cost as well as an increase in the signal-to-noise ratio, as compared to the original single-band algorithm. The MSAD algorithm is developed and three dissimilarity measures that allow multispectral diffusion are suggested. Results for standard anisotropic diffusion and for MSAD with each of the three dissimilarity measures are compared for noisy multi-spectral imagery.