Spatial chromatic contrast sensitivity in natural scenes
Final Report Abstract
We investigated the role of spatial frequencies in the perception of the material properties roughness, thickness, and undulation. In daily life we often have to make rapid visual assessments of the material properties of terrains and objects, e.g., the ground we are walking on or surfaces we are going to touch. Sometimes the best way to estimate the relevant property of a material is to feel it. This, however, is not always feasible. Especially when judgments have to be made rapidly or at greater distance than arms’ length, we have to rely on visual evaluation. In general, we are able to reliably ascribe properties to materials just by visual inspection. These visual judgments are based on the retinal images projected from interactions between illumination-geometry, material-structure and object-shape. The physics of these interactions are too involved for the visual system to estimate material structure by inverse optics. An attractive possibility is that we use heuristics based on information that can be extracted quickly from images. We present a computationally efficient heuristic that is based on the energy at certain bands of spatial frequencies in the images’ amplitude spectra. Starting from the results of a material classification experiment in which we asked observers to classify images of fabrics on the property dimensions soft–rough, flexible–stiff, warm–cold, and water-absorbent–water-repellent, we analyzed the amplitude spectra of the images and found that the material properties roughness, thickness, and undulation could be classified by the energy contained in certain bands of spatial frequencies. The spectra of undulated fabrics contained more energy at low spatial frequencies (0.57– 2.29 cycles per degree) as compared to the spectra of flat fabrics. The spectra of thick and thin fabrics differed in a frequency band slightly higher than the first band (2.29–4.28 cpd), and spectra of rough fabrics contained more energy at middle-frequencies (6.57–15.14 cpd) than the spectra of soft fabrics. We manipulated the appearance of images by varying the energy in the three frequency bands and demonstrated that the perceptions of the three properties vary continuously with the modulations. In addition, we conducted an adaptation experiment to show that the information in the three frequency bands actually influences the perception of the material properties. We found that observers’ judgments of fabric roughness, thickness, and undulation were depressed after adaptation to bandpass filtered noise with the bandpass-filters determined by the corresponding frequency bands. These results suggest that cortical filters for estimating material properties could be constructed by combining the outputs of specific sets of frequency-selective neurons. This study demonstrates that the scale of local structure is critical in the perception of material properties. It provides a model for neurophysiological investigations of material perception, and a method for rapid editing and transfer of material properties for computer graphics and animation.
Publications
- (2011). Detecting animals in natural surroundings: The role of color distributions. Journal of Vision, 11 (11), 360
Jansen, M., Giesel, M., & Zaidi, Q.
(See online at https://doi.org/10.1167/11.11.360) - (2011). Visual perception of material affordances. Journal of Vision, 11 (11), 356
Giesel, M. & Zaidi, Q.
(See online at https://doi.org/10.1167/11.11.356) - (2012).Adaptation reveals frequency band based inferences of material properties. Journal of Vision, 12 (9), 950
Giesel, M. & Zaidi, Q.
(See online at https://doi.org/10.1167/12.9.950)
