Classifier combination is known to generally perform better than each individual classifier by taking into account the complementarity between the input pieces of information. Dempster-Shafer theory is a framework of interest to make such a fusion at the decision level, and allows in addition to handle the conflict that can exist between the classifiers as well as the uncertainty that remains on the sources of information. In this contribution, we present an approach for classifier fusion in the context of large-scale multi-label and multi-modal image classification that improves the classification accuracy. The complexity of calculations is reduced by considering only a subset of the frame of discernment. The classification results on a large dataset of 18,000 images and 99 classes show that the proposed method gives higher performances than of those classifiers separately considered, while keeping tractable computational cost.