The role of natural image statistics in visual memory and recognition

Recent years have seen a resurgent interest in eye movements during natural scene viewing. Aspects of eye movements that are driven by low-level image properties are of particular interest due to their applicability to biologically motivated artificial vision and surveillance systems. In this paper, we report an experiment in which we recorded observers’ eye movements while they viewed calibrated greyscale images of natural scenes. Immediately after viewing each image, observers were shown a test patch and asked to indicate if they believed it was part of the image they had just seen. The test patch was either randomly selected from a different image from the same database or, unbeknownst to the observer, selected from one of the fixations on the image just viewed. We find that several low-level image properties differed significantly relative to the observers’ ability to successfully designate test patches. The goal of the paper was to measure the image properties that facilitate visual memory in a non-cognitive natural setting to test the theory that VSTM and VLTM are implemented iconically rather than as encoded sensory information. Results indicate that the VSTM/VLTM model we tested may need to be re-evaluated since memory for the noncognitive patches used was shown to survive eye movements.