A New Image Denoising Scheme Using Spiking Neuromorphic Systems

Abstract

It has been shown that restricted Boltzmann machines (RBMs) perform efficiently in a variety of applications such as dimensionality reduction, learning and classification. In image processing and computer vision research, image denoising has been used as a preprocessing step to estimate an original image from a noise-contaminated image by suppressing noise. In this work, we propose a new image denoising scheme based on neuromorphic platforms. Image denoising on neuromorphic hardware platforms can have significant advantages from the perspectives of scalability, concurrency and low-power consumption and real-time interaction with the given environment. From the experimental study for the MNIST hand-written digit dataset, we demonstrated that the proposed approach can generally outperform median, Gaussian and Wiener, filtering based denoising schemes, especially for a (relatively) high noise.