Standardizing pre-processing for digital skin lesion image analysis

Pre-processing steps are critical in automated image analysis systems developed to aid in diagnosis of skin lesion images. The main areas of concern include, but are not limited to, hair on the skin, variations in illumination and skin tone, and alignment of successive skin images. These artifacts can partially or completely obstruct a lesion being analyzed causing errors in classification or diagnosis. This paper focuses on an independent quantitative evaluation of an open source hair removal algorithm [1]. The different input parameters to the algorithm were tested to determine their optimal values. Percent error and signal to noise ratio are utilized as the error metrics for the experimental results. Other essential pre-processing steps are considered and provided at the end of this paper.