Exploring Image Quality Assessment from a New Perspective: Pupil Size

TL;DR

This study uses pupillometry to examine how performing an IQA task influences cognitive processing and its relation to image quality. In a two-task experiment, subjects freely observe or rate image quality for 100 LIVE-distorted images while pupil size is tracked; results show IQA engages visual attention and alters recovery dynamics, with higher image quality slowing pupil recovery. The work provides a theoretical basis for incorporating attentional mechanisms into objective IQA models and proposes a new subjective IQA approach that leverages pupil-based authentic impressions. Practically, these insights could improve IQA methods by accounting for cognitive load and attentional allocation during evaluation.

Abstract

This paper explores how the image quality assessment (IQA) task affects the cognitive processes of people from the perspective of pupil size and studies the relationship between pupil size and image quality. Specifically, we first invited subjects to participate in a subjective experiment, which includes two tasks: free observation and IQA. In the free observation task, subjects did not need to perform any action, and they only needed to observe images as they usually do with an album. In the IQA task, subjects were required to score images according to their overall impression of image quality. Then, by analyzing the difference in pupil size between the two tasks, we find that people may activate the visual attention mechanism when evaluating image quality. Meanwhile, we also find that the change in pupil size is closely related to image quality in the IQA task. For future research on IQA, this research can not only provide a theoretical basis for the objective IQA method and promote the development of more effective objective IQA methods, but also provide a new subjective IQA method for collecting the authentic subjective impression of image quality.

Figures (11)

• Figure 1: 20 reference images selected from the LIVE database.
• Figure 2: Distribution of quality of all selected distorted images provided by the LIVE database.
• Figure 3: Experimental procedures. (a) shows the display procedure of stimulus materials in Task 1. (b) shows the display procedure of stimulus materials in Task 2.
• Figure 4: Changes in pupil size with observation time in Task 1 and Task 2. Light colored areas represent standard errors. Note that curves are Gaussian smoothed. The dotted line divides the change in pupil size into two phases: constriction and recovery.
• Figure 5: Changes in pupil size when subjects observed images with different distortion types in Task 1 and Task 2. Light colored areas represent standard errors.