Temporal Statistics of Natural Image Sequences Generated by Movements with Insect Flight Characteristics

Abstract

Many flying insects, such as flies, wasps and bees, pursue a saccadic flight and gaze strategy. This behavioral strategy is thought to separate the translational and rotational components of self-motion and, thereby, to reduce the computational efforts to extract information about the environment from the retinal image flow. Because of the distinguishing dynamic features of this active flight and gaze strategy of insects, the present study analyzes systematically the spatiotemporal statistics of image sequences generated during saccades and intersaccadic intervals in cluttered natural environments. We show that, in general, rotational movements with saccade-like dynamics elicit fluctuations and overall changes in brightness, contrast and spatial frequency of up to two orders of magnitude larger than translational movements at velocities that are characteristic of insects. Distinct changes in image parameters during translations are only caused by nearby objects. Image analysis based on larger patches in the visual field reveals smaller fluctuations in brightness and spatial frequency composition compared to small patches. The temporal structure and extent of these changes in image parameters define the temporal constraints imposed on signal processing performed by the insect visual system under behavioral conditions in natural environments.

Image statistics of the environment surrounding freely behaving hoverflies

“…Our findings are important as they describe some relevant image statistics of the natural environment for different behaviors in freely behaving animals. This has previously been done for human observers (Parraga et al 2000 ; Frazor and Geisler 2006 ), and for flying blowflies (Schwegmann et al 2014b ), but not for hoverflies. These results can in the future be used when designing naturalistic stimuli to understand the neural coding of visual neurons (Dyakova et al 2015 ; Nicholas et al 2018 ) or when quantifying visual behaviors (Goulard et al 2015 ; 2016 ).…”

“…A widely used and well-investigated image statistic is the slope constant of the amplitude spectrum, which can be extracted after doing a Fourier transform of the image (Field 1987 ). On a log–log scale, the rotational average of the amplitude spectra of natural scenes has a shape, with slope constants (alpha) around 1–1.2 (Tolhurst et al 1992 ; Bex and Makous 2002 ; Schwegmann et al 2014b ). When alpha is exactly one, the scene is scale-invariant.…”

“…For example, the amount of man-made structures, the distance between the observer and the scene being depicted, and the level of scene “openness” all have different image statistics, including slope constants (alpha) (Torralba and Oliva 2003 ; McCotter et al 2005 ). Furthermore, within a given scene local image statistics may vary dramatically (Frazor and Geisler 2006 ; Schwegmann et al 2014b ). To understand the visual coding of natural scenes, we thus need to define natural for a particular animal performing a specific behavior.…”

Peripheral Processing Facilitates Optic Flow-Based Depth Perception

“…A panoramic rectangular lattice with square pixels was obtained by unwrapping the panoramic images obtained with the hyperpoloidal mirror system projections. The image sequences obtained in this way were filtered and resized with 2D spatial Gaussian filters to simulate the spatial filtering property of insect photoreceptors, assuming an acceptance angle of 1.64° and interommatidial angular distance of 1.25°, thereby image sequences with panoramic rectangular lattice of 73 × 289 pixel 2 were obtained (for details, see Schwegmann et al, 2014a and the data in Schwegmann et al, 2014c published online).…”

“…Hence, nearness information cannot be extracted unambiguously from EMD responses (Egelhaaf and Borst, 1993 ; Dror et al, 2001 ; Rajesh et al, 2005 ; Straw et al, 2008 ; Meyer et al, 2011 ; O'Carroll et al, 2011 ; Hennig and Egelhaaf, 2012 ). A recent model study suggested that EMD responses to translational optic flow resemble a representation of the contrast-weighted nearness (CwN) to objects in the environment or, in other words, of the contours of nearby objects (Schwegmann et al, 2014a , b ). This conclusion, however, needs to be qualified, because it does not take the dynamic rescaling of local light intensities by the visual system via adaptive processes into account, potentially influencing the extraction of depth information from EMD responses.…”