This paper proposes a novel method to extract melanin and hemoglobin concentrations of human skin, using bilateral decomposition with the knowledge of a multiple layered skin model and absorbance characteristics of major chromophores. Different from state-of-art approaches, the proposed method enables to address highlight and strong shading usually existing in skin color images captured under uncontrolled environment. The derived melanin and hemoglobin indices, directly related to the pathological tissue conditions, tend to be less influenced by external imaging factors and are effective for describing pigmentation distributions. Experiments demonstrate the value of the proposed method for computer-aided diagnosis of different skin diseases. The diagnostic accuracy of melanoma increases by 9-15% for conventional RGB lesion images, compared to techniques using other color descriptors. The discrimination of inflammatory acne and hyperpigmentation reveals acne stage, which would be useful for acne severity evaluation. It is expected that this new method will prove useful for other skin disease analysis.