In image processing, edge detection often makes use of integer-order differentiation operators, especially order I used by the gradient and order 2 by the Laplacian. This paper demonstrates how introducing an edge detector based on non-integer (fractional) differentiation can improve the criterion of thin detection, or detection selectivity in the case of parabolic luminance transitions, and the criterion of immunity to noise, which can be interpreted in term of robustness to noise in general. (C) 2003 Elsevier B.V. All rights reserved.