Image Empirical Mode Decomposition (IEMD) in Dermoscopic Images: Artefact Removal and Lesion Border Detection

R. Fonseca-Pinto, P. Caseiro, and A. Andrade (Portugal)


Artefact removal; Skin cancer; Automatic border detection; Image empirical mode decomposition (IEMD).


Melanocytic lesions are a potential source of skin cancer. Early detection of lesions and the minimization of the high mortality rate associated to skin cancer depend on the awareness of the general public and on adequate training of professionals. Early detection is followed by periodic inspection of lesions, possibly assisted by digital dermatoscopy. One important characteristic of lesion for diagnosis purposes is the shape of the lesion border and its growth rate. Dermatoscopic images of skin lesions are marred by artefacts like hairs and air bubbles producing artificial borders and rendering automatic border detection a challenging task. In this work we present image empirical mode decomposition (IEMD), a new time-frequency signal processing technique adapted to images, to remove common artefacts (air bubbles and hairs) from dermatoscopic images. IEMD can be employed as a preliminary filtering procedure to improve automatic border detection of skin lesions. IEMD shows good performance in artefact removal, and compares favourably with a classical filtering approach in the specific context of lesion border detection.

Important Links:

Go Back