Object perception is inherently multidimensional: information about color, material, texture and shape all guide how we interact with objects. We developed a paradigm that quantifies how two object properties (color and material) combine in object selection. On each experimental trial, observers viewed three blob-shaped objects-the target and two tests-and selected the test that was more similar to the target. Across trials, the target object was fixed, while the tests varied in color (across 7 levels) and material (also 7 levels, yielding 49 possible stimuli). We used an adaptive trial selection procedure (Quest+) to present, on each trial, the stimulus test pair that is most informative of underlying processes that drive selection. We present a novel computational model that allows us to describe observers' selection data in terms of (1) the underlying perceptual stimulus representation and (2) a color-material weight, which quantifies the relative importance of color vs. material in selection. We document large individual differences in the color-material weight across the 12 observers we tested. Furthermore, our analyses reveal limits on how precisely selection data simultaneously constrain perceptual representations and the color-material weight. These limits should guide future efforts towards understanding the multidimensional nature of object perception.

Several domains of neuroscience offer map-like models that link location on the cortical surface to properties of sensory representation. Within cortical visual areas V1, V2, and V3, algebraic transformations can relate position in the visual field to the retinotopic representation on the flattened cortical sheet. A limit to the practical application of this structure-function model is that the cortex, while topologically a two-dimensional surface, is curved. Flattening of the curved surface to a plane unavoidably introduces local geometric distortions that are not accounted for in idealized models. Here, we show that this limitation is overcome by correcting the geometric distortion induced by cortical flattening. We use a mass-spring-damper simulation to create a registration between functional MRI retinotopic mapping data of visual areas V1, V2, and V3 and an algebraic model of retinotopy. This registration is then applied to the flattened cortical surface anatomy to create an anatomical template that is linked to the algebraic retinotopic model. This registered cortical template can be used to accurately predict the location and retinotopic organization of these early visual areas from cortical anatomy alone. Moreover, we show that prediction accuracy remains when extrapolating beyond the range of data used to inform the model, indicating that the registration reflects the retinotopic organization of visual cortex. We provide code for the mass-spring-damper technique, which has general utility for the registration of cortical structure and function beyond the visual cortex.

The first step in the evolution of primate trichromatic color vision was the expression of a third cone class not present in ancestral mammals. This observation motivates a fundamental question about the evolution of any sensory system: how is it possible to detect and exploit the presence of a novel sensory class? We explore this question in the context of primate color vision. We present an unsupervised learning algorithm capable of both detecting the number of spectral cone classes in a retinal mosaic and learning the class of each cone using the inter-cone correlations obtained in response to natural image input. The algorithm's ability to classify cones is in broad agreement with experimental evidence about functional color vision for a wide range of mosaic parameters, including those characterizing dichromacy, typical trichromacy, anomalous trichromacy, and possible tetrachromacy.

Cones with peak sensitivity to light at long (L), medium (M) and short (S) wavelengths are unequal in number on the human retina: S cones are rare (<10%) while increasing in fraction from center to periphery, and the L/M cone proportions are highly variable between individuals. What optical properties of the eye, and statistical properties of natural scenes, might drive this organization? We found that the spatial-chromatic structure of natural scenes was largely symmetric between the L, M and S sensitivity bands. Given this symmetry, short wavelength attenuation by ocular media gave L/M cones a modest signal-to-noise advantage, which was amplified, especially in the denser central retina, by long-wavelength accommodation of the lens. Meanwhile, total information represented by the cone mosaic remained relatively insensitive to L/M proportions. Thus, the observed cone array design along with a long-wavelength accommodated lens provides a selective advantage: it is maximally informative.

Color vision supports two distinct visual functions: discrimination and constancy. Discrimination requires that the visual response to distinct objects within a scene be different. Constancy requires that the visual response to any object be the same across scenes. Across changes in scene, adaptation can improve discrimination by optimizing the use of the available response range. Similarly, adaptation can improve constancy by stabilizing the visual response to any fixed object across changes in illumination. Can common mechanisms of adaptation achieve these two goals simultaneously? We develop a theoretical framework for answering this question and present several example calculations. In the examples studied, the answer is largely yes when the change of scene consists of a change in illumination and considerably less so when the change of scene consists of a change in the statistical ensemble of surface reflectances in the environment.

Demosaicing is an important part of the image-processing chain for many digital color cameras. The demosaicing operation converts a raw image acquired with a single sensor array, overlaid with a color filter array, into a full-color image. In this paper, we report the results of two perceptual experiments that compare the perceptual quality of the output of different demosaicing algorithms. In the first experiment, we found that a Bayesian demosaicing algorithm produced the most preferred images. Detailed examination of the data, however indicated that the good performance of this algorithm was at least in part due to the fact that it sharpened the images while it demosaiced them. In a second experiment, we silenced image sharpness as a factor by applying a sharpening algorithm to the output of each demosaicing algorithm. The optimal amount of sharpening to be applied to each image was chosen using the results of a preliminary experiment. Once sharpness was equated in this way, an algorithm developed by Freeman based on bilinear interpolation combined with median filtering, gave the best results. An analysis of our data suggests that our perceptual results cannot be easily predicted using an image metric.

This paper describes the design and performance of an image capture simulator. The general model underlying the simulator assumes that the image capture device contains multiple classes of sensors with different spectral sensitivities and that each sensor responds in a known way to irradiance over most of its operating range. The input to the simulator is a set of narrow-band images of the scene taken with a custom-designed hyperspectral camera system. The parameters for the simulator are the number of sensor classes, the sensor spectral sensitivities, the noise statistics and number of quantization levels for each sensor class, the spatial arrangement of the sensors and the exposure duration. The output of the simulator is the raw image data that would have been acquired by the simulated image capture device. To test the simulator, we acquired images of the same scene both with the hyperspectral camera and with a calibrated Kodak DCS-200 digital color camera. We used the simulator to predict the DCS-200 output from the hyperspectral data. The agreement between simulated and acquired images validated the image capture response model and our simulator implementation. We believe the simulator will provide a useful tool for understanding the effect of varying the design parameters of an image capture device.

Vision algorithms are often developed in a Bayesian framework. Two estimators are commonly used: maximum a posteriori (MAP), and minimum mean squared error (MMSE). We argue that neither is appropriate for perception problems. The MAP estimator makes insufficient use of structure in the posterior probability. The squared error penalty of the MMSE estimator does not reflect typical penalties. We describe a new estimator, which we call maximum local mass (MLM) [10, 26, 65], which integrates the local probability density. The MLM method is sensitive to local structure of the posterior probability, which MAP is not. The new method uses an optimality criterion that is appropriate for perception tasks: it finds the most probable approximately correct answer. For the case of low observation noise, we provide an efficient approximation. We apply this new estimator to color constancy. An unknown illuminant falls on surfaces of unknown colors. We seek to estimate both the illuminant spectrum and the surface spectra from photosensor responses which depend on the product of the unknown spectra. In simulations, we show that the MLM method performs better than the MAP estimator, and better than two standard color constancy algorithms. The MLM method may prove useful in other vision problems as well.>