Using models of the Human Visual System in the design of stack filters for the enhancement of color images

A technique is developed for utilizing models of the Human Visual System to improve the design of filters for the enhancement of color images. The technique uses an image fidelity measure based on models of the human visual system — such as the Visible Differences Predictor (VDP) — in a nested loop training algorithm. In the inner loop of the algorithm, a stack filter is trained under a Weighted Mean Absolute Error (WMAE) Criterion to remove noise. In the outer loop, the VDP is used to train the wights in the WMAE criterion to ensure that the filter to which the algorithm converges is one that produces output images that are as visually satisfying as possible. The stack filters resulting from this VDP-driven, WMAE approach perform much better than filters trained under the standard mean absolute error criterion. This fact is demonstrated with color images. The robustness of these trained filters to variations in both the image and the noise is discussed.