In order to reduce the influence of mixed pixel on dictionary, the author has constituted principal component analysis (PCA)united sparse dictionary from the first principal component extracted with sparse dictionary of panchromatic image and unmiximg image by the online dictionary learning algorithm and PCA. The sparse dictionary can include multi-spectral image and high-spatial resolution image features, while considering the mixed pixel problem. The sparse coefficients of panchromatic and multi-spectral images are calculated using PCA united sparse dictionary and orthogonal matching pursuit(OMP) algorithm, then the sparse coefficients of fusion image are calculated using nonnegative matrix factor(NMF) fusion algorithm, thus reconstructing fusion image. In consideration of the root mean square error of the reconstructed image and the limitation of computing, research on the dictionary matrix size shows that the final matrix size of sparse dictionary is 64×480. An analysis of five quantitative assessment indexes demonstrates that more texture details and multi-spectral information can be obtained by the proposed fusion than by united sparse dictionary NMF fusion, wavelet fusion and PCA fusion. The proposed method can obtain better fusion result.