In this paper, a restoration approach for hyperspectral images with two observations of specific spatial and spectral resolutions is presented. This proposed approach is based on a multi-band image restoration scheme employing Expectation-Maximization (EM) algorithm, in which both observation models are used for restoration. Simulation experimental results illustrate that compared to existing restoration approaches, in which only one observation model is utilized while the other observation is treated as auxiliary, the proposed restoration approach is capable of producing improved restored results.