Ashley Alfred / Mathematics / Faculty Mentor: Jianzhong Su
Hyperspectral Microscopy Imaging (HMI) combines hyperspectral imaging (HSI) with microscopy to capture both spatial and spectral details at a microscopic level, enabling detailed chemical analysis. However, accurately identifying bacterial species remains challenging due to dataset variations and inherent biases in hyperspectral data analysis. Key steps in HMI analysis include using unbiased data, selecting optimal feature extraction techniques, and applying methods such as principal component analysis (PCA) and machine learning classifiers. This work explores compressed sensing (CS) as a sparsity-driven, regularization-based approach for image enhancement. By using sparse signal representations, CS facilitates better separation of signal and noise, while regularization ensures stability and accuracy. While CS has been widely applied in image reconstruction, its potential for feature enhancement remains underexplored. Traditional CS techniques prioritize smooth recovery by minimizing the total variation (TV) norm, which can suppress fine details. Instead, we propose minimizing the L1 norm to enhance specific image features, particularly edges, without enforcing overall smoothness. To evaluate the effectiveness of this approach, we use peak signal-to-noise ratio (PSNR) and structural similarity index (SSIM) to quantify image similarity. By successfully amplifying rough features, this method aims to improve bacterial classification accuracy in HMI data analysis.bits.
Leave a Reply