Skin-Sensitive Automatic Color Correction
Nachlieli, Hila; Bergman, Ruth; Greig, Darryl; Staelin, Carl; Oicherman, Boris; Ruckenstein, Gitit; Shaked, Doron
Keyword(s): skin detection, face detection, adaptive enhancement, object dependent enhancement, image probability map, color correction, sharpening
Abstract: One of the most important aspects of a good image is that the subject looks good. It is therefore our objective, when building an automatic image enhancement system, to optimize the appearance of the image subject(s). Moreover, people often have specific preferences regarding the look of different objects. Grass, for example, is considered prettier when green and with sharp texture, while skin is expected to be smooth, and, preferably, not green. An image enhancement process that involves such object dependent considerations requires the guidance of a human supervisor able to identify image content and adapt the enhancement parameters for each content object accordingly. We propose to automate content sensitive image enhancement to correspond to human preferences regarding the appearance of people in an image. We do this by detecting the skin tones of specific people in a given image. Based on these learned skin tones, we generate a fuzzy object map representing skin areas in the image. This map is used to automatically tune parameters of the enhancement algorithms. We present an adaptive skin color correction algorithm. We demonstrate the effectiveness of our algorithms in both a laboratory setting and in a production system.
Additional Publication Information: Submitted to SIGGRAPH 2009, New Orleans.
External Posting Date: February 6, 2009 [Fulltext]. Approved for External Publication
Internal Posting Date: February 6, 2009 [Fulltext]