Real-time Human Corneal Images Processing Analysis of Placido Disk Corneal Topography Using Extenics

摘要

A sub-pixel image edge detection method based on polar coordinates is proposed in order to accurately detect images obtained from corneal topographic instrument based on Placido disk. The corneal image obtained by the corneal topographic instrument has the following characteristics: 1. Real time. The image acquisition video is set to 30 frames/sec, so the image processing algorithm should have the millisecond level of computing performance; 2. Quality is not stable. Because of the influence of photographing environment, machine and human entry factors, the images of cornea will have various kinds of noises. So we should propose a stable and effective preprocessing algorithm. 3. Image offset. Due to the operation characteristics of such corneal topographic system, the consecutive images’ centers are unavoidable drifted during the process of image acquisition. So it is necessary to match the same serial target pictures. According to the above characteristics, based on the theory of extension and matter-element data structure, this paper efficiently structured captured images and uses two-dimensional Gaussian smoothing function to preprocess the data of all kinds of noises, including real human eyelashes removing and uses mixed Canbel (combining the canny and Sobel algorithms) and regional three-polynomial fitting method to effectively detect the sub-pixel edges of corneal images which we derive. Based on the experiments of standard simulation eye and real human eye, it is proved that the algorithm presented in this paper has good robustness and detection accuracy (the refractive error distributions of standard simulation eyes with different curvatures are less than 0.25D), which meets the operation requirements of actual systems used in industry. So our algorithm can be used in such corneal topographer in order to help detection of various eye diseases.