Fig 1.
A 68 facial landmark detector pretrained from the iBUG 300-W dataset.
Each number represents a predefined feature on the face. For eye extraction, we only considered the index numbers from 37 to 48 on the eye region.
Fig 2.
The flowchart of the proposed method.
A frontal facial image is sent to the face detection model to identify the face region and the detected face region is subsequently used to extract the eye region through the facial landmark detector. Otsu’s binarization and the HSV color model are applied to the extracted eye region image, and the results from two methods are used to form a new image. The pixel points located at the limbus are sampled and used to estimate the pupil center by the LSM. Finally, the positional similarity of the iris on both eyes is computed for strabismus screening. (Note that eye images displayed in this paper were converted to gray to protect private information. The proposed method actually uses RGB images as the input to perform strabismus screening).
Fig 3.
Eye extraction using the 68 facial landmark detector.
(a) Result of right eye area extraction. (b) Result of left eye area extraction. Dots represent the landmarks surrounding the eye region.
Fig 4.
Image processing on eye regions.
(a)-(c) show the results of Otsu’s binarization, the HSV color model, and the combination of two methods on the left eye region. (d)-(f) are the results of the right eye region.
Fig 5.
The result of pixel-point sampling on the left and right eye limbus.
(a) Right eye region. (b) Left eye region. Dots represent the sampled pixel points on the limbus.
Fig 6.
Estimation of pupil center coordinate by the LSM.
(a) Right eye region. (b) Left eye region. Dots represent the estimated pupil center.
Fig 7.
Distance estimation between the pupil center and canthus landmarks.
(a) Right eye region. (b) Left eye region. R1 and R2 are the distance measurements within the right eye using the estimated pupil center and the coordinates of the medial and lateral canthus. L1 and L2 are the distance measurements within the left eye.
Table 1.
The sample mean and sample standard deviation of the positional similarity estimates of normal and strabismus images.
Fig 8.
Distribution of positional similarity of normal and strabismus images.
Fig 9.
The positional similarity estimates of 10 normal and strabismus images.
(a) The normal images. (b) The strabismus images. The numbers on both sides represent the positional similarity estimates.