A novel segmentation algorithm has been developed for acne vulgaris lesion. Instead of applying conventional color-spaces, mapping of RGB color pixels to visible electromagnetic spectrum is performed as it provides finer discrimination between shades of red brown and purple. RGB color of the image is approximately mapped to the equivalent visible spectrum from 380 nm to 720 nm with the steps of 1 nm. Average spectral models for different class of colors are first selected from various images. The algorithm classifies the lesion and non-lesion pixels based on Mahalanobis distance from the spectral models. Candidate pixel is labeled as an entity whose spectral model is least distant. © 2013 American Scientific Publishers All rights reserved. All rights reserved.