We introduce a Matryoshka approach based on the Iterated Inside Out (IIO) algorithm for solving optimal transport problems in image processing. IIO alternates between improving non-basic variables with negative reduced costs (inside phase) and restoring basic feasibility (outside phase). We propose an image-processing variant that efficiently identifies variables with negative reduced costs through two specialized procedures. This variant is embedded in the Matryoshka approach, which constructs a hierarchy of transportation problems at progressively finer image resolutions, solving each optimally and using its solution as the starting point for the next level. This multiscale strategy produces high-quality initial solutions and accelerates resolution. Experiments on DOTmark instances show that Matryoshka-IIO outperforms the best existing exact solvers.