Memristive crossbar array-based frameworks for image analysis and classification

Abstract

This thesis explores application of an yttrium oxide (Y₂O₃)-based memristive crossbar array (MCA) model, MCA developed through a dual ion beam sputtering system, for high cyclic stability in resistive switching applications. The experimentally obtained data from the fabricated MCA was validated against an analytical MCA-based model, showing excellent alignment with experimental results. Utilizing this validated model, we applied it to biomedical image processing, specifically in analysing computed tomography (CT) and magnetic resonance imaging (MRI) images, through a two-dimensional image decomposition technique. By employing varying decomposition levels and threshold values, we evaluated reconstructed image quality through metrics such as peak signal-to-noise ratio (PSNR), structural similarity index (SSIM), and mean square error (MSE). In our analysis, MRI and CT scan images exhibited compression ratios of 21.01% and 47.81% using Haar and 18.82% and 46.05% with biorthogonal wavelets. Brightness analysis showed significant improvements in image quality, with increases of 103.72% for CT scans and 18.59% for MRI images using Haar wavelets. These findings underscore the potential of the MCA-based model for image compression, facilitating reduced computation times and storage requirements in biomedical engineering.