Image binarization method based on classification framework

Abstract

The invention belongs to the technical field of computer image processing, and discloses an optical coherence tomography scanning eye image binaryzation method and system based on a classification framework. The method comprises the steps of: carrying out designated simple small region division on a small number of optical coherence tomography scanning eye sample images, and marking a region of interest and a background; then extracting features of each target region by using a Gabor template, splicing the features into a feature vector, performing dimension reduction on the extracted features, and learning feature vector mapping subspaces; introducing a support vector machine learning classifier model; processing a classification result image by adopting an algorithm; and finally, obtaining a binary image. The method has the advantages that the small sample binaryzation problem is solved, the optical coherence tomography scanning eye image binaryzation method based on a specified N-neighborhood target region classification framework is provided, the small samples can be fully utilized to train a support vector machine classification model, and the self-adaptive binaryzation threshold of the image samples can be obtained through learning.

Description

technical field [0001] The invention belongs to the technical field of computer image processing, in particular to an image binarization method based on a classification framework. Background technique [0002] Binarization is a fundamental task in optical coherence tomography (OCT) image analysis, and it has been widely used for both simple quantification of area and more complex analysis of blood vessels. In recent years, a great deal of research has focused on the quantitative analysis of octagonal images of retinal vessels and choroidal capillaries. OCT can obtain images of living tissues with micron resolution, and has been successfully applied in medical diagnosis, such as diagnostic ophthalmology and other human internal organs. OCT is a non-invasive imaging technique. [0003] Machine learning methods have been greatly developed in biomedical image processing. These methods have been widely used in biomedical image reconstruction, image denoising, and disease clas...

AI Technical Summary

Problems solved by technology

[0017] Aiming at the problems existing in the existing algorithms, the present invention proposes an image binarization method based on the classification framework, which can overcome the problem of limited contrast in the image, and can use small samples to learn the classification model

Images