Automatic image processing system for beautifying human faces

Abstract

In this paper,we propose an automatic image processing system to beautify human faces in frontal-parallel color images. Although most of image processing packages provide functions to beautify color images, a few of them, at least based on our knowledge, are specific for beautifying human faces. By using these functions, the processed face images become unreal. For example, they will remove most of natural edges around some special regions, such as eyes and mouth. Therefore, the proposed system only processed the regions of faces. To make the processed face region smoother, our system treats the regions of eyes/mouth and the rest of face region differently. By using different methods to smooth these two types of regions, we can keep almost all the natural edges around eyes and mouth , but remove wrinkles and spots on the rest of faces. The process of our system ia as follows. At beginning, we convert the RGB color space into YCbCr space so as to segment face region from scene background based on the value range of the skin color proposed by H.A. Rowly, etc. Within the face region, the system uses the chain-code to get the eye region and the mouth region. For the eye and mouth regions, we adjust the image pixels by pixels; the rest of pixels are justified by block base. To evaluate the performance of our system, we compare our system with the tool?cleanSkinFX, which can be found at the Web site http://www.mediachance.com. Our system is outperforming.