Detection and analysis of hijab based on visual feature of neck and hair

In the recent decade facial feature and hijab detection is an important task for human computer interaction. To detect and recognize the women in all aspects hijab detection is a vital issue as hijab covers many important features in face such as ear, head and face's corner portions. In this regard, a hijab detection framework is proposed in this work. For that, initially, the face and eyes regions are detected by using efficient Viola-Jones cascade classifier. After that, the facial and it's connected surrounding face color region (FCR) is extracted by YCbCr color model. As hair is located upper surrounding portion of the FCR, the left, right and top FCR border points are estimated by utilizing eyes bounding box's left, right height's center point and bounding box center point. This eyes bounding box center point is also used to estimate the bottom neck point (NP) from FCR and chin point (CP) from bottom facial region. Furthermore, the y coordinate of the NP and CP is used to confirm the neck cover. And the left, right and top FCR border points are used to confirm the hair cover. Finally, based on the neck and hair cover the hijab is detected. The proposed method is tested using various hijab as well as non-hijab images and results are presented to demonstrate the efficiency and effectiveness.