Fig. 3: Effects of fixational drift magnitude on reconstruction quality. | Nature Communications

Fig. 3: Effects of fixational drift magnitude on reconstruction quality.

From: Fixational eye movements enhance the precision of visual information transmitted by the primate retina

Fig. 3

a Image reconstruction performance for three preparations (delineated with different colors) as a function of the magnitude of eye movements simulated during the stimulus presentation, for joint-LNBRC-dCNN (solid line), known-LNBRC-dCNN (dashed line), and zero-LNBRC-dCNN (dotted line). The magnitude of drift eye movements was quantified as the standard deviation in the eye position occurring during each trial. The error bars in every panel correspond to the standard error of mean reconstruction quality. In all preparations, reconstruction quality for joint-LNBRC-dCNN and known-LNBRC-dCNN increased with eye position jitter, up to (but not including) the largest eye movements evaluated. Reconstructions for zero-LNBRC-dCNN were less accurate than both known-LNBRC-dCNN and joint-LNBRC-dCNN and further decreased with increasing eye movements. b Eye position estimation error as a function of the magnitude of movement, for the same experimental preparations. When eye movements were ignored (zero-LNBRC-dCNN, dotted line), the error in estimated eye position increased linearly, as expected with a 2D Brownian motion. When eye movements were jointly estimated (joint-LNBRC-dCNN; solid lines), the error increased, but more gradually. c Parasol-only (solid line) and midget-only (dashed line) joint-LNBRC-dCNN image reconstruction performance as a function of the magnitude of movement, for the same experimental preparations. In all preparations, reconstruction quality increased with increasing magnitudes of fixational drift for both parasol-only and midget-only reconstructions. Midget-only reconstructions had systematically better quality than parasol-only reconstructions in all preparations. d Parasol-only (solid line) and midget-only (dashed line) eye position estimation error, for the same experimental preparations. For both parasol-only and midget-only reconstructions, the eye position estimation error increased more slowly than if eye movements were ignored (dotted line).

Back to article page