Kaiki Chiu¹ (), Emily Lo¹, Quinn O'Connor¹, Michelle R. Greene¹; ¹Barnard College, Columbia University, Department of Psychology

Human visual processing is rapid and accurate even in the face of complex real-world scenes. A common framework for explaining this feat posits that the brain creates efficient representations of visual inputs by capitalizing on statistical redundancies (Attnaeve, 1954). A key implication of this framework is the testable hypothesis that images with higher redundancy (i.e., lower information content) undergo more efficient processing than less redundant (i.e., higher information content) counterparts. However, quantifying the information content of images may be an intractable challenge (Chandler & Field, 2007). In this study, we propose a novel approach by focusing on the relative information content within scenes, which we will estimate by measuring the relative compression efficacy of these images through widely accessible algorithms, such as JPEG and PNG. Specifically, our rationale is that more easily compressed images are likely to possess greater redundancy and thus less information. We amassed a database comprising ~1000 photographs of everyday scenes in RAW image format. We compressed each image in PNG (lossless) format and compared the file size differences between the original and compressed images. From this dataset, we selected the 100 most and 100 least compressible images. Observers (N=39) performed a rapid detection task in which they distinguished between scene images and 1/f noise (SOA: ~80 ms, with a dynamic pattern mask). Observers had higher scene detection sensitivity for the highly compressible images (d'=3.15 vs 2.88, p<0.001), indicating that images with lower relative information content were processed more easily, supporting our hypothesis. Our findings demonstrate that visual processing efficiency is influenced by the relative information content of scenes, as measured by compression algorithms. This aligns with the efficient coding hypothesis, suggesting that the visual system achieves efficiency by exploiting environmental regularities.

Acknowledgements: NSF CAREER 2240815 to MRG.