Detecting high-level and low-level properties in visual images and visual percepts