Answers are grounded in this paper and cite their sources.
Visual scene recognition and modeling rely heavily on textures, which encapsulate an object's material properties across varying scales, lighting, and viewing angles, influenced by factors such as object shape, lighting conditions, and observational distance. However, evaluating the success of texture synthesis or recognition currently requires costly and impractical psychophysical human evaluations.This study aims to bridge this gap by comprehensively comparing various texture quality and similarity metrics against psychophysically derived human rankings. Utilizing a specialized texture quality benchmark, we evaluate metrics originating from both texture-specific research (TQA) and general image quality assessment (IQA) that do not require pixelwise correspondence. Our findings highlight the superior performance of two multispectral measures derived from the adaptive recursively evaluated Markovian descriptive model. While these objective numerical alternatives show promise for virtual prototyping and scene recognition, the results suggest that current state-of-the-art methods still require refinement to ensure universal reliability across diverse computer vision and pattern recognition domains. Replacing manual evaluation with these efficient, statistically validated measures is fundamental for advancing physically accurate texture modeling and synthesis.
No takes yet. Share an insight, caveat, or question.
Haindl et al. (2026) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: