Measuring and Predicting Object Importance.

How important is a particular object in a photograph of a complex scene? We propose a definition of importance and present two methods for measuring object importance from human observers. Using this ground truth, we fit a function for predicting the importance of each object directly from a segmented image; our function combines a large number of object-related and image-related features. We validate our importance predictions on 2,841 objects and find that the most important objects may be identified automatically. We find that object position and size are particularly informative, while a popular measure of saliency is not.

Tunable color filters based on metal-insulator-metal resonators.

We report a method for filtering white light into individual colors using metal-insulator-metal resonators. The resonators are designed to support photonic modes at visible frequencies, and dispersion relations are developed for realistic experimental configurations. Experimental results indicate that passive Ag/Si(3)N(4)/Au resonators exhibit color filtering across the entire visible spectrum. Full field electromagnetic simulations were performed on active resonators for which the resonator length was varied from 1-3 microm and the output slit depth was systematically varied throughout the thickness of the dielectric layer. These resonators are shown to filter colors based on interference between the optical modes within the dielectric layer. By careful design of the output coupling, the resonator can selectively couple to intensity maxima of different photonic modes and, as a result, preferentially select any of the primary colors. We also illustrate how refractive index modulation in metal-insulator-metal resonators can yield actively tunable color filters. Simulations using lithium niobate as the dielectric layer and the top and bottom Ag layers as electrodes, indicate that the output color can be tuned over the visible spectrum with an applied field.

Objects predict fixations better than early saliency.

Humans move their eyes while looking at scenes and pictures. Eye movements correlate with shifts in attention and are thought to be a consequence of optimal resource allocation for high-level tasks such as visual recognition. Models of attention, such as "saliency maps," are often built on the assumption that "early" features (color, contrast, orientation, motion, and so forth) drive attention directly. We explore an alternative hypothesis: Observers attend to "interesting" objects. To test this hypothesis, we measure the eye position of human observers while they inspect photographs of common natural scenes. Our observers perform different tasks: artistic evaluation, analysis of content, and search. Immediately after each presentation, our observers are asked to name objects they saw. Weighted with recall frequency, these objects predict fixations in individual images better than early saliency, irrespective of task. Also, saliency combined with object positions predicts which objects are frequently named. This suggests that early saliency has only an indirect effect on attention, acting through recognized objects. Consequently, rather than treating attention as mere preprocessing step for object recognition, models of both need to be integrated.