Automatic acne detection for medical treatment

In this paper, an effective image processing technique has been studied to develop a system of acne detection. The focus is on binary thresholding applied to facial images with various types, shapes or amounts of acne. A typical image on a cheek has been usually used to take in the experiment which results are markings on acnes automatically, this method is more effective than manual counting by a typical dermatologist. An input image is first taken into gray and special color model from regular red-green-blue version, this work is for comparison and revelation of region of interest. The regions of interest are mostly acnes but it can contain some noise. Then binary threshold has been used to clearly show and can remove the noise easier from regions of acne.

A box shape is constructed to mark the final result of acne detection, so that correct and incorrect detected acne can be displayed and calculated for a system efficiency. Experimental output can be summarized that accuracy, precision and sensitivity are quite fluctuated depending on color, shape and lighting condition of acnes in theimage.