In this paper, hyperspectral and multispectral image fusion approach, based on spectral decomposition and neighborhood relation, is proposed. Firstly, endmembers and abundances are extracted from hyperspectral image by using spectral decomposition. Then, an abundance map for neighbouring pixels and endmembers is extracted from the multispectral image. Finally, a high spatial resolution hyperspectral image is produced by using endmembers, neighbouring pixels and abundance map. The proposed method is tested on real hyperspectral images and experimental results show that it outperforms similar fusion algorithms in the literature.