Spatial-Spectral Feature Extraction with Local Covariance Matrix from Hyperspectral Images through Hybrid Parallelization

This paper presents the optimization and hybrid parallelization of a spatial–spectral Feature Extraction (FE) method from hyperspectral images (HSI) using local covariance matrix (CM) representation, exploiting parallelism through multicore manycore processors. The aim is to evaluate performance parallel versions this innovative algorithm that characterizes information prior classification when conducting feature extraction. HSI first projected into subspace, maximum noise fraction method. Then, for each test pixel, its most similar neighbors are clustered cosine distance measurement. result used calculate CM with nondiagonal entry characterizing correlation between different spectral bands. Such matrices represent features fed Support Vector Machine (SVM) classification. To optimize successive process, new version original MATLAB code has been developed C language. serial serves as baseline in OpenMP CUDA. Performance analysis conducted publicly available datasets, confirming our implementation ensures quality while significantly reducing involved processing times.