Automatic phase-based edge detection of corneal Sheimpflug images

In this paper, a new method is proposed for automatic edge detection of the front and back corneal contours in the frame of Sheimpflug images. First, we enhance the contrast and reduce noises including speckle and eyelashes as the preprocessing procedure. Then we use the method of phase symmetry and phase asymmetry to get energy images. Finally, we trace the center line of the cornea and then the upper and lower curves. We validated the proposed segmentation method on a series of videos collected by the department of Chinese PLA General Hospital. The results shows that the proposed method provides equivalent performance as the manual method and better performance than built-in method of the machine. In addition, it is demonstrated that the proposed method is more robust to noise and provides more accurate segmentation results under the presence of eyelashes than other traditional edge detection methods.