Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Handcrafted Feature-Assisted One-Class Learning for Artist Authentication in Historical Drawings

Hassan Ugail, Jan Ritch-Frel, Irina Matuzava

TL;DR

The paper addresses the challenge of authenticating historical drawings when reference data are scarce and stylistic cues are primarily line-based. It introduces a verification-based scheme using ten artist-specific one-class autoencoders trained on five handcrafted features (Fourier energy, Shannon entropy, global contrast, GLCM homogeneity, and box-counting fractal dimension). The study evaluates 10 artists with 20 training images each and 9 test images per artist, totaling 900 verification decisions, and reports pooled TAR of 83.3% and FAR of 9.5% at a fixed operating point, along with Wilson confidence intervals to address small-sample uncertainty. A detailed analysis of per-artist performance and false-accept pathways reveals both strong discriminability and artist-related confusability, underscoring the method's value as a reproducible, quantitative aid in data-scarce attribution tasks that complements connoisseurship. The approach offers a practical, interpretable tool for heritage contexts and can be extended to other domains with limited training data and line-focused media.

Abstract

Authentication and attribution of works on paper remain persistent challenges in cultural heritage, particularly when the available reference corpus is small and stylistic cues are primarily expressed through line and limited tonal variation. We present a verification-based computational framework for historical drawing authentication using one-class autoencoders trained on a compact set of interpretable handcrafted features. Ten artist-specific verifiers are trained using authenticated sketches from the Metropolitan Museum of Art open-access collection, the Ashmolean Collections Catalogue, the Morgan Library and Museum, the Royal Collection Trust (UK), the Victoria and Albert Museum Collections, and an online catalogue of the Casa Buonarroti collection and evaluated under a biometric-style protocol with genuine and impostor trials. Feature vectors comprise Fourier-domain energy, Shannon entropy, global contrast, GLCM-based homogeneity, and a box-counting estimate of fractal complexity. Across 900 verification decisions (90 genuine and 810 impostor trials), the pooled system achieves a True Acceptance Rate of 83.3% with a False Acceptance Rate of 9.5% at the chosen operating point. Performance varies substantially by artist, with near-zero false acceptance for some verifiers and elevated confusability for others. A pairwise attribution of false accepts indicates structured error pathways consistent with stylistic proximity and shared drawing conventions, whilst also motivating tighter control of digitisation artefacts and threshold calibration. The proposed methodology is designed to complement, rather than replace, connoisseurship by providing reproducible, quantitative evidence suitable for data-scarce settings common in historical sketch attribution.

Handcrafted Feature-Assisted One-Class Learning for Artist Authentication in Historical Drawings

TL;DR

The paper addresses the challenge of authenticating historical drawings when reference data are scarce and stylistic cues are primarily line-based. It introduces a verification-based scheme using ten artist-specific one-class autoencoders trained on five handcrafted features (Fourier energy, Shannon entropy, global contrast, GLCM homogeneity, and box-counting fractal dimension). The study evaluates 10 artists with 20 training images each and 9 test images per artist, totaling 900 verification decisions, and reports pooled TAR of 83.3% and FAR of 9.5% at a fixed operating point, along with Wilson confidence intervals to address small-sample uncertainty. A detailed analysis of per-artist performance and false-accept pathways reveals both strong discriminability and artist-related confusability, underscoring the method's value as a reproducible, quantitative aid in data-scarce attribution tasks that complements connoisseurship. The approach offers a practical, interpretable tool for heritage contexts and can be extended to other domains with limited training data and line-focused media.

Abstract

Authentication and attribution of works on paper remain persistent challenges in cultural heritage, particularly when the available reference corpus is small and stylistic cues are primarily expressed through line and limited tonal variation. We present a verification-based computational framework for historical drawing authentication using one-class autoencoders trained on a compact set of interpretable handcrafted features. Ten artist-specific verifiers are trained using authenticated sketches from the Metropolitan Museum of Art open-access collection, the Ashmolean Collections Catalogue, the Morgan Library and Museum, the Royal Collection Trust (UK), the Victoria and Albert Museum Collections, and an online catalogue of the Casa Buonarroti collection and evaluated under a biometric-style protocol with genuine and impostor trials. Feature vectors comprise Fourier-domain energy, Shannon entropy, global contrast, GLCM-based homogeneity, and a box-counting estimate of fractal complexity. Across 900 verification decisions (90 genuine and 810 impostor trials), the pooled system achieves a True Acceptance Rate of 83.3% with a False Acceptance Rate of 9.5% at the chosen operating point. Performance varies substantially by artist, with near-zero false acceptance for some verifiers and elevated confusability for others. A pairwise attribution of false accepts indicates structured error pathways consistent with stylistic proximity and shared drawing conventions, whilst also motivating tighter control of digitisation artefacts and threshold calibration. The proposed methodology is designed to complement, rather than replace, connoisseurship by providing reproducible, quantitative evidence suitable for data-scarce settings common in historical sketch attribution.
Paper Structure (36 sections, 18 equations, 4 figures, 7 tables, 1 algorithm)

This paper contains 36 sections, 18 equations, 4 figures, 7 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Context and motivation
  3. Data scarcity and the rationale for one-class verification
  4. The role of handcrafted features in data-scarce settings
  5. Scope and contributions
  6. Related Work
  7. Machine learning in art authentication
  8. Autoencoders and anomaly detection
  9. Feature engineering for art analysis
  10. Biometric verification frameworks
  11. Materials and Methods
  12. Dataset and curation
  13. Preprocessing
  14. Feature extraction
  15. Fourier energy
  16. ...and 21 more sections

Figures (4)

  • Figure 1: Sample sketches from the dataset showing stylistic diversity across the ten artists.
  • Figure 2: Diagramatic representation of the feedforward autoencoder architecture for one-class sketch verification.
  • Figure 3: Pooled confusion matrix across all 900 verification decisions. Actual genuine vs impostor on rows; predicted genuine vs non-genuine on columns; counts TP=75, FN=15, FP=77, TN=733.
  • Figure 4: Operating points for each artist-specific verifier plotted in FAR--TAR space. The pooled operating point is shown for reference (FAR=0.095, TAR=0.833).