Image Feature Extraction Using Matrix Calculus Techniques: Applications in Texture Analysis and Pattern Recognition.

Abstract

Abstract: Feature extraction constitutes an essential stage in digital image processing, as it enables the transformation of high-dimensional visual information into compact and discriminative representations. This article presents a scientific study focused on the use of matrix calculus techniques for feature extraction in digital images, with particular emphasis on texture analysis and pattern recognition. Images are modeled as numerical matrices, which allows for the systematic application of linear algebra tools such as linear transformations, matrix decompositions, and eigenvalue-based statistical methods. Classical techniques are analyzed, including the Discrete Fourier Transform, the Discrete Cosine Transform, Singular Value Decomposition, and Principal Component Analysis, as well as statistical descriptors derived from co-occurrence matrices. The study highlights the relevance of these methods due to their mathematical interpretability, computational efficiency, and applicability in pattern recognition systems.