[For part I, see ibid. 29, No. 1, 182–193 (1992; Zbl 0746.65091).]
The authors study a class of nonlinear parabolic integro-differential equations for image processing. The diffusion term is modelled in such a way, that the dependent variable diffuses in the direction orthogonal to its gradient but not in all directions. Thereby the dependent variable can be made smooth near an “edge”, with a minimal smoothing of the edge.
A stable algorithm is then proposed for image restoration. It is based on the “mean curvature motion” equation. Application of the solution is persuasively demonstrated for several cases.