EUVIMEDEUVIMED entdecken
Crossref

On the Theory and Application of Singular Value Decomposition in Image Processing and Data Analysis

Gayane Ghazaryan, Artashes Ghazaryan

Mathematical Problems of Computer Science · 01.06.2026 · 65 · 30-48

DOI: 10.51408/1963-0144 ↗

Kurz gefasst

Fachartikel in der Zeitschrift Mathematical Problems of Computer Science, online erschienen am 01.06.2026. Verfasst von Gayane Ghazaryan und 1 weiteren Autor:innen.

AbstractBibliografische AngabenZitierenVerwandte Studien

Abstract (Originalsprache)

Singular Value Decomposition (SVD) is a fundamental factorization method central to image processing, dimensionality reduction, and numerical linear algebra. Its practical strength lies in truncating small singular values, thereby reducing storage while preserving the essential structure. The key challenge is choosing the truncation rank, which determines the trade-off between accuracy, efficiency, and stability. We introduce an adaptive rank selection method that combines cumulative-energy thresholds with automated elbow detection, yielding a principled alternative to ad hoc rules. The framework is validated on image compression, denoising, and principal component analysis (PCA), with benchmarks against eigenvalue decomposition (EVD) and QR factorization. The results show that the adaptive rule improves fidelity and reproducibility, especially for noisy or ill-conditioned data. All code, figures, and tables are released as open source, ensuring that the experiments can be reproduced in a clean environment. This unifies theory, automation, and reproducibility, reaffirming that SVD is both mathematically optimal and practically versatile for contemporary data analysis.

Zitieren

APA

Ghazaryan, G., & Ghazaryan, A. (2026). On the Theory and Application of Singular Value Decomposition in Image Processing and Data Analysis. Mathematical Problems of Computer Science, 65, 30-48. https://doi.org/10.51408/1963-0144

Vancouver

Ghazaryan G, Ghazaryan A. On the Theory and Application of Singular Value Decomposition in Image Processing and Data Analysis. Mathematical Problems of Computer Science. 2026;65:30-48. doi:10.51408/1963-0144

BibTeX
@article{Ghazaryan2026,
  author = {Ghazaryan, Gayane and Ghazaryan, Artashes},
  title = {On the Theory and Application of Singular Value Decomposition in Image Processing and Data Analysis},
  journal = {Mathematical Problems of Computer Science},
  year = {2026},
  volume = {65},
  pages = {30--48},
  doi = {10.51408/1963-0144}
}
Kostenlos starten →