A machine vision-based image feature extraction method for anterior segment tomographic images, comprising: first, carrying out grayscale histogram statistics on anterior segment tomographic images collected by a camera; removing images which could not possibly be anterior segment images; then carrying out grayscale normalization according to a histogram to reduce the impact of ambient light on imaging quality; then by means of a K-mean clustering algorithm, performing rough segmentation, wherein cornea, iris and crystalline lens areas may be segmented out; then carrying out blob analysis, and screening out non-anterior segment images according to the position relationship and shape information of each area; and then carrying out fine boundary tracking in a fixed direction on the basis of the rough boundary of each area so as to obtain accurate profiles of the cornea, iris and crystalline lens, and provide reliable basic data for the subsequent finding of anterior segment clinical parameters.
展开▼