Time: Monday, October 16, 2023, 18:30

Location:  Lecture Hall 537, Building 3, Hainayuan, Zijingang Campus, Zhejiang University

The memorability-attention nexus: how scene memorability influences visual search

Speaker: Qi Li

Researcher, Okayama University, Japan

Qi Li earned her bachelor's degree from the School of Journalism at Fudan University. During her master's and doctoral studies at Kyoto University, she conducted research on visual working memory and attention under Professor Sokitsu Takaki. She has subsequently engaged in research and education activities at Kyoto University, the University of Tokyo, and Okayama University. Her current research projects primarily focus on visual working memory, visual attention, image memorability, and data judgment.

Abstract

The saying a picture is worth a thousand words highlights that although scene images are complex combinations of vast visual information, our visual system can process image information extremely rapidly. However, not all images are equally memorable—some remain vivid long after viewing, while others fade quickly from memory. This inherent property of images is termed memorability. In the field of engineering, image memorability is receiving increasing attention. Recent research suggests that image memorability is an independent, fundamental characteristic. It is not determined by an image's hue, brightness, or saturation, nor is it related to statistical features like the number of objects, nor does it depend on subjective perceptions such as aesthetics or interestingness. So, how might image memorability relate to attention? In cognitive psychology, extensive empirical studies have demonstrated that memory and attention are closely linked; information that is attended to receives more processing resources and has a greater chance of entering the memory system. To explore the relationship between image memorability and attention, we employed a visual search paradigm. By comparing search efficiency in two conditions—searching for a highly memorable image among less memorable images versus searching for a less memorable image among highly memorable images—we discovered a novel search asymmetry phenomenon induced by image memorability. Specifically, search efficiency significantly decreased when looking for a less memorable image among highly memorable images, compared to the reverse scenario. We propose that highly memorable images attract attention more readily, and once attention is captured by these images, it becomes harder to disengage, thereby reducing the efficiency of searching for less memorable images. Our research reveals the close relationship between image memorability and visual attention, while also demonstrating that the search asymmetry paradigm can serve as an effective method for studying image memorability.