Image segmentation enhances discrimination of motion in visual noise

Abstract

The primate visual system uses form cues-such as hue, contrast polarity, luminance, and texture-to segment complex retinal images into the constituent objects of the visual scene. We investigated whether segmentation of dynamic images on the basis of hue, luminance contrast polarity, or luminance contrast amplitude aids discrimination of motion direction. Human subjects viewed dynamic displays of randomly positioned dots, in which a variable proportion of the dots moved in the same direction at the same speed ("signal" dots) while the remaining dots were randomly displaced ("noise" dots). In agreement with previous reports, we observed a reliable relationship between the strength of the motion signal and subjects' ability to discriminate motion direction, enabling the measurement of thresholds for direction discrimination. When signal dots had a different luminance contrast amplitude than noise dots, direction discrimination performance was directly related to the relative contrast of the signal dots, demonstrating the importance of matching the perceived contrast amplitude of signal and noise tokens when testing the effects of segmentation by other cues. When Michelson luminance contrast was matched, distinguishing signal from noise dots by hue or by luminance contrast polarity strongly improved direction discrimination, lowering thresholds by an average factor of five. These results reveal a strong influence of form cues on motion processing in the human visual system, and suggest that segmentation on the basis of form cues occurs prior to motion processing.