An improved memristive model driven cellular neural networks for highly efficient advanced image processing

Abstract This study presents an advanced window function designed to address the limitations inherent in current memristive models, which often encounter issues such as boundary effects, boundary-locking, high complexity, and inflexibility. Memristors, which emulate synaptic behavior in neural circuits, rely on window functions to accurately model the nonlinear doping drift phenomenon. The proposed window function enhances adjustability and flexibility, aiming to mitigate these prevalent issues. Through MATLAB simulations, this research examines the volt-ampere characteristics of memristors and benchmarks the performance of the improved window function against traditional models. Additionally, the enhanced function is applied to image processing tasks, including edge extraction and denoising, in conjunction with a cellular neural network. The results demonstrate that the improved model effectively resolves the boundary-locking issue and exhibits superior hysteresis behavior. Furthermore, the memristor-based cellular neural network outperforms existing algorithms in image processing, as evidenced by improved peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM). These findings underscore the effectiveness and advantages of the proposed window function in both memristor characterization and practical image processing applications.