Sub-pixel Edge Detection And Parameters Calculation

The feature of corneal surface can be analyzed by corneal topography.Doctor can detect the surface of human corneal with corneal topography and obtain useful information which will be used to guide the treatment,the result of detection can be used to identify and detect corneal disease.Corneal topography apparatus is an instrument used to assess the shape of the surface of the cornea.This thesis studies the image processing of the image of the human cornea based on the Placido disk.The main contents of the study include image edge detection,ring extraction,sub-pixel edge detection,circle fitting,eliminating the noise point and calculating the center and radius.In order to find the edge point at pixel level,canny edge detection was used to detect the image.Rings of the image were extracted by computing connected components.Concerning the low accuracy in localization in traditional edge detection algorithm and high precision edge are demanded in the work,a sub-pixel edge detection algorithm based on quadratic function interpolation method was proposed.Least square circle fitting method was used to calculate the center of each ring.As the centers may exist noise point,kmeans was used to cluster the centers and eliminate the noise point,however it is all know that kmeans suffers from initial condition effects,so a method was proposed to determine the initial condition of the kmeans algorithm.Experiment results show that the proposed method have a good clustering results.Rings with different radius have different weights when calculating the concentric circle center.In order to reduce the influence of noise,least absolute deviation was used to calculate the radius of the ring.By extracting the region of interest,not only ensure the result of the experiment,but also quicken the implementary speed of the procedure.High accuracy edge can be extracted and obtain effective parameters by using the present method in the paper.