Aesthetic Index for Art Paintings Using Visual Features
2025 IEEE 5th International Conference on Smart Information Systems and Technologies (SIST), Astana
Yerdauit Torekhan, Nurdaulet Altynbekov, Pakizar Shamoi
School of Information Technology and Engineering, Kazakh-British Technical University
Abstract
In the digital era, assessing image aesthetics has become essential for various applications, including photography ranking, content recommendation, and human-computer interaction. However, due to the subjective nature of human perception, defining and quantifying aesthetics remains a challenging task in computational aesthetics. This paper introduces a novel approach to evaluating the aesthetic quality of paintings based on three fundamental features: symmetry, simplicity, and color harmony. By applying computational techniques to measure these attributes, we propose an aesthetic index that provides an objective assessment of artistic composition. We conduct an experiment on a large-scale dataset of paintings, “Best Artworks of All Time,” to analyze the effectiveness of our method. The results reveal that Kazimir Malevich, Paul Gauguin, and Joan Miró are among the artists with the highest average aesthetic scores. Additionally, we identify the top 10 most aesthetically pleasing artworks, with paintings from Malevich, Rublev, and Klee ranking among the highest. The average computed aesthetic score across all artworks is 71.74. Our results show that basic visual characteristics can be used to measure a painting’s aesthetic appeal. The paper methodology offers a means of assessing and contrasting artworks that can be used to build aesthetic composition trends.
Method
Three visual features
Rather than training an end-to-end aesthetic predictor, the index is built from three interpretable properties measured directly on the painting. Symmetry captures the balance of the composition about its axes. Simplicity captures how much visual complexity the canvas carries. Color harmony captures how the palette relates to itself. Each is computed independently, so a painting's score can always be traced back to the feature that produced it.
Combining into a single index
The three feature scores are aggregated into one aesthetic index, which makes the ranking of an entire collection possible in a single pass. Because the components stay separate until the final step, the index remains interpretable: two paintings with the same score can be distinguished by which feature carried it.
Results
Ranking Best Artworks of All Time
Applied across the "Best Artworks of All Time" dataset, the index produces a per-artist ranking by mean aesthetic score, with an average of 71.74 across all artworks. Kazimir Malevich, Paul Gauguin and Joan Miró come out among the highest-scoring artists; at the level of individual paintings, the top ten is led by works from Malevich, Rublev and Klee. The pattern is consistent with what the simplicity and colour-harmony terms reward — flat, deliberately restricted palettes over dense figurative detail.
Key points
- Aesthetic quality reduced to three measurable features — symmetry, simplicity, color harmony.
- Interpretable by construction: every score traces back to its components.
- Mean aesthetic score of 71.74 across the Best Artworks of All Time collection.
Cite
@inproceedings{torekhan2025aesthetic,
title = {Aesthetic Index for Art Paintings Using Visual Features},
author = {Torekhan, Yerdauit and Altynbekov, Nurdaulet and Shamoi, Pakizar},
booktitle = {2025 IEEE 5th International Conference on Smart Information
Systems and Technologies (SIST)},
year = {2025},
doi = {10.1109/SIST61657.2025.11139166}
}