Exploring Kolmogorov-Arnold networks for realistic image sharpness assessment

TL;DR

The paper investigates the applicability of Kolmogorov-Arnold Networks (KANs) for score prediction in blind image sharpness assessment, introducing TaylorKAN as a Taylor-series-based variant. It evaluates six KAN variants alongside SVR and MLP on four realistic databases using 15 mid-level and 2048 high-level features, detailing implementation and performance characteristics. Across datasets, KANs are generally competitive or superior to baseline models, with TaylorKAN excelling when mid-level features are used, while high-level features yield mixed results due to high dimensionality. The study highlights the potential of KANs in quality assessment tasks, underscores the value of feature selection and integration with deep representations, and outlines directions for improving representation power and generalization. Overall, it establishes KANs as a promising tool for score prediction in image quality evaluation and motivates further exploration of diverse basis functions and post-processing strategies.

Abstract

Score prediction is crucial in evaluating realistic image sharpness based on collected informative features. Recently, Kolmogorov-Arnold networks (KANs) have been developed and witnessed remarkable success in data fitting. This study introduces the Taylor series-based KAN (TaylorKAN). Then, different KANs are explored in four realistic image databases (BID2011, CID2013, CLIVE, and KonIQ-10k) to predict the scores by using 15 mid-level features and 2048 high-level features. Compared to support vector regression, results show that KANs are generally competitive or superior, and TaylorKAN is the best one when mid-level features are used. This is the first study to investigate KANs on image quality assessment that sheds some light on how to select and further improve KANs in related tasks.