Adaptive technique for matching the spectral response in skin lesions' images

Abstract

The suggested technique is a subsequent stage for data obtaining from diffuse reflectance spectra and images of diseased tissue with a final aim of skin cancer diagnostics. Our previous work allows us to extract patterns for some types of skin cancer, as a ratio between spectra, obtained from healthy and diseased tissue in the range of 380 – 780 nm region. The authenticity of the patterns depends on the tested point into the area of lesion, and the resulting diagnose could also be fixed with some probability. In this work, two adaptations are implemented to localize pixels of the image lesion, where the reflectance spectrum corresponds to pattern. First adapts the standard to the personal patient and second – translates the spectrum white point basis to the relative white point of the image. Since the reflectance spectra and the image pixels are regarding to different white points, a correction of the compared colours is needed. The latest is done using a standard method for chromatic adaptation. The technique follows the steps below:–Calculation the colorimetric XYZ parameters for the initial white point, fixed by reflectance spectrum from healthy tissue;–Calculation the XYZ parameters for the distant white point on the base of image of nondiseased tissue;–Transformation the XYZ parameters for the test-spectrum by obtained matrix;–Finding the RGB values of the XYZ parameters for the test-spectrum according sRGB;Finally, the pixels of the lesion's image, corresponding to colour from the test-spectrum and particular diagnostic pattern are marked with a specific colour.