Human image quality judgment relies on low dimensional perceptual spaces
Multidimensional Observer Model and Perceptual Dimensions of Human Image Quality Assessment
Computer Vision and Pattern Recognition
Summary
Judging how good an image looks is important for daily life and for machines that work with images. The authors created a new model that mimics how humans mentally compare images by imagining them as points in a simple space with only a few key features. Their model matches real human choices and shows that people use very few basic ways to judge image quality, even though images themselves are very complex. It also reveals that people use different kinds of information depending on whether they are focused on simple details or overall meaning in images.
What this means in practice
- •For image processing engineers: Improve image quality metrics used in compression or enhancement by incorporating low-dimensional perceptual representations reflecting human judgment.
- •For game developers: Design visual content that prioritizes dimensions most relevant to human quality perception for better player experience with limited resources.
Authors
Sheng Zhao, Weikai Lin, Yuhao Zhu
Abstract
Judging image quality is not only ecologically relevant to everyday human tasks, but also underpins many machine vision tasks such as image generation. This paper proposes a framework to understand the inherent perceptual space underlying image quality judgment in humans. We propose a multi-dimensional observer model that represents images as distributions in a latent perceptual space and that models human judgment as comparing noisy samples. Being constrained by neural representations in the primate ventral stream and fit to large-scale behavioral data, the model enables analysis of perceptual structure while matching the predictive power of existing metrics. Using this model, we find that the perceptual spaces needed to account for image quality judgment in humans are extremely low-dimensional compared to the image space even when considering its sparsity. The exact structure of the space (e.g., dimensionalities, information encoded) varies between low-level and high-level quality judgments, suggesting that, despite a shared retinal encoding in the beginning, humans selectively construct task-dependent perceptual spaces in visual decision making.