Image enhancement and noise suppression for Optical Coherence Tomography images based on variational image decomposition and gaussian mixture model

Abstract

Optical Coherence Tomography (OCT) images often suffer from low contrast and severe speckle noise. For such need, this work presents a new image enhancement and noise suppression method for OCT images. A new variational image decomposition (VID) model TV-Hilbert-Curvelet is proposed in order to decompose the OCT image into three part: structure, background and noise. And the gaussian mixture model (GMM) is used to generate one binary mask template from the background part, the multi scale Retinex (MSR) is used to enhance the structure part. Experimental results show that the proposed method can enhance the image structure and suppress speckle noise well, and three quality indexes are used to verify the experimental results of the proposed method.