A generalized framework for histogram-based similarity measures is presented and applied to the image-enhancement task in digital subtraction angiography (DSA). The class of differentiable, strictly convex weighting functions is identified as suitable weightings of histograms for measuring the degree of clustering that goes along with registration. With respect to computation time, the energy similarity-measure is the function of choice for the registration of mask and contrast image prior to subtraction. The registration success for the automated procedure is compared with a manually shift-corrected image pair of the head.