Towards Detection of Glasses in Facial Images

Eyeglasses Lens Contour Extraction from Facial Images Using an Efficient Shape Description

“…Eyeglasses detection was pioneered by Jiang et al [3], who proposed six measures for determining the presence of eyeglasses in facial images based on edge information from two regions around the eyes. This method requires that the eyes are detected in advance and does not provide any information regarding the localization or the shape of the eyeglasses.…”

“…Jing et al [5] extended the methods for detecting the presence of eyeglasses presented in [3], and extracted the eyeglasses based on deformable contours, using geometrical and edge features. The final position of the eyeglasses is determined by means of dynamic programming.…”

“…The information regarding the interpupillary distance is used to delimit the region in which the eyeglasses are likely to be found, in a similar manner as presented in [3]. Figure 6 depicts the computed search area of the eyeglasses based on the eyes position and the approximation of the interpupillary distance.…”

“…For example, the eyeglasses region of interest could be robustly extracted using Active Appearance Models ([24]) or Constrained Local Models ([25]). We chose, however, to compute this zone heuristically, as presented in [3,5], because it is very simple, does not require a training phase and complies with the conditions we imposed for the hypotheses generation space. This region must be large enough to comprise the contours of the eyeglasses and sufficiently narrow, such as the search space to be limited.…”

“…The first step of the algorithm is to determine the approximate region of the eyes; based on the position of the eyes, the eyeglasses search region is determined to be similar to the method presented in [3]. Next, we use a Monte Carlo search method to determine the approximate position and size of the eyeglasses.…”