Hyperspectral imaging (HSI), with its ability to capture information across hundreds of contiguous spectral bands, offers a rich source of data for detailed thematic mapping and classification. However, the high dimensionality inherent in hyperspectral data poses significant challenges for efficient analysis, often necessitating the application of dimensionality reduction techniques. In this study, we introduce a novel pixel-based dimensionality reduction method tailored for hyperspectral imagery. The method involves segmenting each pixel's spectral signature curve (SSC) into equal-length intervals. Two new mathematical indices one based on the integral of each segment and the other on the L2-norm are then used to extract representative features from each segment, producing a compact and informative feature vector. Unlike traditional approaches such as Principal Component Analysis (PCA), the proposed technique preserves the sequential structure of the SSC and can be independently and concurrently applied to all pixels. This property makes the method highly suitable for parallel processing architectures. To evaluate its performance, the method was tested on two benchmark hyperspectral datasets in agricultural domains and was compared against PCA and wavelet-based techniques. Experimental results demonstrate that the proposed method outperforms the baselines in terms of classification accuracy and computational efficiency, making it a promising tool for real-time hyperspectral data analysis.