Reducing noise is an important task for improving visual and automated assessment of medical images. In this talk I will primarily use retinal OCT images as a target application; OCT images suffer from considerable speckle noise. Traditional image/signal processing methods only offer moderate speckle reduction; deep learning methods can be more effective but require substantial training data, which may not be readily available. I will describe our novel self-fusion method that offers effective speckle reduction comparable to deep learning methods, but without any external training data. I will present qualitative and quantitative results in a variety of datasets from fovea and optic nerve head regions, with varying SNR values for input images. I will also talk about our recent work to combine deep learning and self-fusion for excellent denoising performance and computational efficiency.
Posted by: Cuong Ly