Visual statistics of aquatic environments in the natural habitats of zebrafish

Abstract

According to the efficient coding hypothesis, animals’ visual systems are adapted to exploit the regularities of natural environments so as to encode maximal sensory information with minimal metabolic cost. The visual regularities of natural environments can be characterized empirically through large image or video datasets. While common visual features have been discovered in the study of these datasets, different environments and contexts also contain reliable statistical differences. Thus, to evaluate the efficiency of a model animal’s visual system, one must characterize the statistics of the animal’s specific habitat. Zebrafish have become a popular model in visual neuroscience due to their amenability to advanced research tools and their diverse set of visually-guided behaviors. However, little is known about the spatiotemporal features of the habitats where zebrafish reside. To address this gap, we collected and analyzed a video dataset of the aquatic environments in northeastern India where zebrafish are native. We obtained this dataset from an omnidirectional high-frame-rate video camera, which was mounted to a robot that captured stationary viewing and simulated swimming behaviors. Similar to previous work on both atmospheric and aquatic natural images, we found that the spatial power spectra of these environments fell off approximately exponentially as a function of frequency. However, we found that the temporal power spectra were less monotonic than expected from atmospheric counterparts. Reduced monotonicity can be attributed to power bumps at mid-range temporal frequencies associated with high-contrast and dynamic rippling patterns underwater. Our results suggest that the demands placed on the zebrafish visual system, particularly with respect to estimating environmental and self motion, differ substantially from the demands placed on terrestrial animals. Based on these results, we will explore the implications of the particular patterns of temporal frequency and speed in the zebrafish environment for driving visually-guided behaviors that require motion detection.