Feature synergy enhances saliency of figure-ground segregation

Texture areas which differ from the surround in more than one local feature are moreeasily detected. Recently, it has been shown that a low-level summary statistic, net contrastenergy, predicts this double-cue advantage, suggesting integration at an early stageof image analysis. Here we asked whether there is also a double-cue advantage in moreadvanced shape-from-texture judgments beyond figure-ground segregation. In targetdetection and a more complex shape discrimination task d’ sensitivity was adjusted tothe same baseline levels for targets with feature contrast in orientation and spatial frequencycues, and performance was measured for double-cue targets. Contrary to earlierresults reported for simpler shape discriminations we found a weaker double-cue advantagein the more complex shape identification task. Particularly, double-cue sensitivitywas notably lower than the algebraic sum of the single-cue sensitivities, a level whichwas consistently reached in the detection task. Control measurements with high featuredifferences between target and background showed perfect detection performance,both with single cues and their combination. Shape identification performance, however,saturated at levels between 83% − 90% accuracy, depending on the feature condition,indicating that enhanced target saliency is insufficient for extracting unique shape representationsfrom the texture stimuli. The findings corroborate that cue summation enhancesfigure-ground segregation in cluttered images, while there is no evidence that italso augments more specific processes involved in the recognition of 2D texture shapes.