On the granularity of summative kernels (Q835191)

In a digital signal processing context, summative kernels are widely used to model the impulse response and to ensure a smooth interplay between continuous and discrete domains. In the paper, summative kernels mean the probability densities over a fixed domain \(\Omega\). Based on their transformation to maxitive kernels (possibilistic densities), a new index called granularity is proposed and examine. Several experimental comparisons stressing the applicability of the gragularity index have been performed and they have highlighted the fact that, when two summative kernels with different shapes have the same granularity, their behaviours in typical kernel-based signal processing applications are quite close (in fact, they are closer than two summative kernels having the same Shannon index).