An Approach to Detecting Human Nipple Regions Using Color Map

Abstract

In this paper, we suggest a new method to detect female nipple areas, representing harmfulness of input images, by employing a color nipple map and geometrical features. The proposed method first detects human skin color regions from an image. It then defines a color nipple map using the unique characteristics that human nipples have and extracts candidate nipple areas by applying the defined map to the input image. Subsequently, our method selects only real nipple regions after filtering out non-nipple ones by applying geometrical features to candidate nipple areas. Experimental results show that the suggested nipple map and geometrical feature-based algorithm can accurately detect human nipple areas through various experiments.