Research Of Glasses Removal Method In Face Recognition

With the rapid development of computer Internet technology,the face recognition technology increasingly has received attention of the research community.Although in recent years many improvements have been made in the area of face recognition,the recognition results usually are affected by facial expressions,gestures,locations,scales and other obstructions.As one of the most common occluding object in facial images,eyeglasses have a great influence on subsequent processing and the face recognition performance.Thus it can be seen the research of the removal of glasses is so important.This paper mainly focuses on the removal of glasses.Because the dark-color-frame eyeglass and the “white-color-frame” whos color has large difference with skin all have strong structure and texture character,it would cause performance loss in face recognition.Therefore it is necessary to remove eyeglass before face recognition.The main contribution of this paper is as follows:1?The detection of occluded region is precondition for eyeglasses removal.Threshold segmentation based on half threshold and the minimum variance filtering was performed to determine whether there were glasses and obtain the occluded region of eyeglasses.Firstly,in order to eliminate the effect of the eyes in the face,eyes were replaced by oval and its inner region.Then half threshold segmentation was performed to take out the obvious areas including eyes and obtain obvious area containing glasses.In view of the particularity of shape of glasses and in order to reduce subtle influence around glasses to get the occluded region of eyeglasses accurately,threshold segmentation based on the minimum variance filtering with multiple variance filter template was performed.Finally,occluded region of eyeglasses was got.The method can reduce interference and noise around glasses.2?The occlusion area of the glasses is large area in the face,and the inpainting algorithms based on texture synthesis usually is used to inpaint large damaged area.The improved method based on the algorithm of Criminisi is used to remove glasses.The calculation of priority takes the position that refer to eyes into account and it also narrows searching scope of the best matched patch according to the middle process of the detection of eyeglasses.The experimental results show that the improved method compared with the Criminisi algorithm has a better removal effect,and leave less traces of glasses.The recognition rate improves significantly.3?Due to differences in close skin area is small,the smaller the distance between the compensation sphere compensation the better effect is.Fan linear interpolation compensation method based on the minimum distance search is proposed to remove glasses.The original compensation sphere was got according to the detection of occlusion region.Then,in order to reduce influence of glasses shape and noise,the adjustment of compensation sphere was performed.The algorithm of searching the match compensation sphere according to the principle of minimum distance was performed to obtain the best match compensation sphere.Finally,the algorithm of the fan linear interpolation compensation were performed to get the face image without glasses.In the process,finding the best match compensation sphere quickly and determining the size of the fan-shaped area is illustrated.The method is simple to realize.the visual effect is good,and the recognition rate improves significantly.