Workflow from Scientific Research

Simplified overview of the iCID optimization. First we initialize the reference images R_{L*} and separate the grid points into subsets G_{LAB}_{,i}. Then we optimize the grid points by repeating the following procedure for a fixed number of times: We sequentially iterate over all subsets G_{LAB}_{,i} and for each grid point x G_{LAB}_{,i} we search a new tonal value z. Based on its current tonal value B(x) we consider a set of alternative tonal values and modify the transform B[x,z]. For each of these modified B[x,z] we compute the metric M using the input LAB_{L*} and corresponding references R_{L*}. The tonal value z that produces the smallest metric value becomes B(x). Note that the grid points in a subset G_{LAB}_{,i} can be optimized in parallel, then the metric is only evaluated in the partition P[x] of each grid point x G_{LAB}_{,i}.

DOI