An Effective Hybrid Harris Hawk Optimization for Facial Skin Segmentation Under Varying Illumination

The segmentation of facial color images is an essential step for facial analysis purposes such as face recognition, identification, and planning of facial reconstruction surgeries. The varying illumination has a notable effect on it. One of the applications of facial skin segmentation is contour extraction in the analysis of facial plastic surgeries, which is a challenging problem under varying illumination. Therefore, in this paper, a modified version of the Fuzzy c-Means (MFCM) algorithm with adding varying illumination parameter is presented to segment frontal and profile facial color images. MFCM algorithm is sensitive to the initial value and may cause this algorithm to fall in a local minimum. In this paper, to overcome the mentioned problems, we proposed a hybrid optimization method, which combines Grey Wolf Optimization (GWO) and Harris Hawk Optimization (HHO). The main goal of using GWO is to improve the exploration phase in HHO. Also, the same weight coefficient is used for all three alpha, beta, and delta wolves. The ranking of wolves for selecting these coefficients is not considered. To improve the location update, weight coefficient is updated based on the rank of each wolf. Experimental results demonstrate that the proposed algorithm has high efficiency and is robust to the varying illumination effect in the segmentation of facial color images. Also, it shows that the proposed algorithm has a suitable performance in facial skin segmentation compared to other image segmentation methods.

Keywords:

Facial Color Image , Facial Skin Segmentation , Fuzzy c-means Clustering , Grey Wolf Optimization , Harris Hawk Optimization , Varying Illumination Correction