Sub-pixel mapping is technique used to obtain the spatial distribution of different classes at the sub-pixel scale by transforming fraction images to a classification map with a higher resolution. Traditional sub-pixel mapping algorithms only utilize a low-resolution image, the information of which is not enough to obtain a high-resolution land-cover map. The accuracy of sub-pixel mapping can be improved by incorporating auxiliary datasets, such as multiple shifted images in the same area, to provide more sub-pixel land-cover information. In this paper, a sub-pixel mapping framework based on a maximum a posteriori (MAP) model is proposed to utilize the complementary information of multiple shifted images. In the proposed framework, the sub-pixel mapping problem is transformed to a regularization problem, and the MAP model is used to regularize the sub-pixel mapping problem to be well-posed by adding some prior information, such as a Laplacian model. The proposed algorithm was compared with a traditional sub-pixel mapping algorithm based on a single image, and another multiple shifted images based sub-pixel mapping method, using both synthetic and real hyperspectral images. Experimental results demonstrated that the proposed approach outperforms the traditional sub-pixel mapping algorithms, and hence provides an effective option to improve the accuracy of sub-pixel mapping for hyperspectral imagery.