SUPER-RESOLUTION AND WAVELET-BASED FBG SIGNAL PROCESSING FOR NON-INVASIVE DERMATOLOGICAL ANALYSIS

Abstract

In recent years, dermatological diagnostics have increasingly relied on advanced imaging technologies and signal processing techniques to enhance early detection and analysis of skin anomalies. This paper proposes a novel framework that synergistically combines Super-Resolution Generative Adversarial Networks (SR-GANs) with wavelet-based Fiber Bragg Grating (FBG) sensor signal emulation for high-precision skin area analysis. The methodology begins by employing SR-GAN to enhance the quality of low-resolution skin images, transforming them into high-resolution outputs that preserve fine texture and edge details critical for clinical evaluation. These super-resolved images are then subjected to simulated digital acquisition and quantization through an Analog-to-Digital Converter (ADC), followed by reconstruction via a Digital-to-Analog Converter (DAC) to mimic real-world sensor capture conditions. Grayscale conversion and adaptive thresholding are applied to estimate affected skin regions with high accuracy. To emulate real time biomedical sensing, the system models six distinct FBG sensors, each with unique wavelength ranges, using advanced discrete wavelet decomposition techniques. By converting 2D skin images into 1D signal vectors through statistical projection, wavelet transforms capture localized signal features relevant to biological variability. Each sensor simulates a wavelength-specific intensity profile, including Gaussian-modulated cosine variations and randomized shifts, to represent realistic physiological noise and environmental perturbations. First and second-order derivatives of intensity with respect to wavelength are computed to enhance signal interpretability and facilitate peak detection, aiding in accurate localization of skin anomalies.