Abstract: The assessment of posiratic lesions on human skin by human observers is not objective or consistent, and the estimates typically involve too considerable variation. To facilitate the evaluation of psoriatic patients, and objective assessment of involved skin surface is needed.
In this work, a segmentation scheme was developed to divide the skin surface into non-involved and involved regions. Based on the bi-modality of small subimages, a variable thresholding scheme was generated. To take both local and global impacts into account, an iterative method with a variable window size was developed. Multiple passes improve the classification of pixels. A number of post-processing steps are applied. These steps operate on a small 3x3 neighborhood of the centre pixel. The number of false segments is reduced, defects are removed from the image and smoother transition is obtained.