OCT image denoising method based on dense connection and generative adversarial network

Abstract

The invention discloses an OCT (Optical Coherence Tomography) image denoising method based on dense connection and a generative adversarial network, which belongs to the technical field of image restoration, and adopts a multi-frame registration mode to synthesize a reference image according to the characteristics of noise randomness, so that the network can learn a mapping relationship between anoise image and the reference image; the noise synthesis step can effectively expand the diversity of speckle noise and synthesize new sample data; dense fusion is carried out by using multi-scale features of the network, and repeated utilization and transmission of effective features are enhanced by using fewer parameters; the overall perception quality of the image is ensured by adopting an adversarial generative network mode; the trained generative model can directly process noise OCT images with any resolution, has high speed and performance, and has high use value in clinic.

Description

technical field [0001] The invention belongs to the technical field of image restoration, and in particular relates to an OCT image denoising method based on dense connection and generation confrontation network. Background technique [0002] Optical coherence tomography (Optical Coherence Tomography, OCT) is a non-invasive, reproducible, 3D fundus tissue imaging technology, which has many applications in ophthalmology. Quantitative analysis of OCT images of eye tissue is helpful It is used in the diagnosis of glaucoma, age-related macular degeneration, diabetic retinal edema and other eye diseases by doctors. OCT is a high-resolution imaging technology that is easily affected by the objective acquisition environment, and the imaging beam is scattered and coherently superimposed by intraocular tissues multiple times, forming speckle noise, which is a common phenomenon in OCT images. Several properties of speckle noise make its removal challenging. Different from ordinary a...

AI Technical Summary

Problems solved by technology

Although super-resolution, anti-aliasing and denoising problems are to restore the details of the image as much as possible, the goals of these tasks are different, and there are great differences between different types of training images. For the denoising problem of OCT images ,not effectively

Although in natural images and other medical fields, deep learning based on adversarial networks has been researched and applied in image restoration, but currently there is no denoising method based on deep learning applied in the field of OCT.

Images