References

Automatic segmentation of skin lesions (e.g. naevi, melanoma) from surrounding healthy skin tissue is essential for designing effective and efficient computer-based methods for diagnosis and prognosis of melanocytic diseases. Fully automatic tumor segmentation is hard due to variability of several factors:- tumor morphology (shape, size, structure, occluding hair)- intrinsic image attributes (skin pigmentation, color, contrast)- acquisition parameters (imaging devices, image resolution, lens deformation