This paper introduces four new dynamic dictionary learning methods to sparse representation based hyperspectral resolution enhancement. The impact of the type and size of the dynamic dictionary to the reconstruction quality is investigated for the recently proposed sparse representation based multiresolution image fusion method J-SparseFI-HM. Low resolution hyperspectral and high resolution multispectral input images are simulated from recently acquired airborne HySpex data. Experiments reveal that fusion products can be substantially improved by changing the dictionary type from the currently used nearest neighbor selection to a modified dissimilarity based dynamic dictionary.