Discrimination of the Fourier amplitude spectrum slope across the visual field.

Bruno Richard
Psychology Ph.D.
Concordia University

This doctoral research is an inquiry into human visual sensitivity to consistent features of natural scenes (e.g., Mountains, Beaches, Forests, etc.). Throughout the past decade, there have been consolidated efforts to relate the characteristics of the content found in natural scenes to the processing strategies of the early visual system. More specifically, it has been hypothesized that visual encoding is tuned to the spatial relationships (e.g., regularities) present in the content of the visual world. We can investigate this hypothesis by empirically measuring the extent to which manipulations of the regularities of natural stimuli can impact visually driven performance. The spatial relationship this research is predominately interested in is the Fourier amplitude spectrum, also called the power spectrum of natural scenes. We can measure the Fourier amplitude spectrum via a two-dimensional discrete Fourier transform, which decomposes any complex waveform (i.e., natural images are complex waveforms of luminance) into a sum of sinusoidal waveforms, and represented as a function of spatial frequency (e.g., scale or detail) and orientation. Such representations are often referred to the amplitude and phase of the waveforms, which represent the amplitude (i.e., height of the wave or energy) at each spatial frequency and the orientation at each spatial frequency respectively. Via multiple measurements of the Fourier amplitude in natural scenes, we find that the amplitude peaks at low spatial frequencies (i.e., coarse scale) and falls with increasing frequency at a constant rate (α). Across natural scenes, the average α range varies between 0.59 and 1.66, and subsequent experimental investigations have found that human are quite sensitive to α values within this range, and less so to more extreme values.

Though it is true that previous investigations have demonstrated that humans are sensitive to the spatial characteristics of natural scenes (e.g., the Fourier amplitude spectrum slope, α), the stimuli used have always remained relatively small, and therefore only stimulated the center part of our visual field, and therefore retina, a region referred to as the fovea. Although the fovea is the area of high acuity in our visual field, and therefore most encoding efforts accomplished by the visual system stem from information collected at the fovea, it is not the only region receiving input from the surrounding environment. To properly measure human discrimination of the spatial regularities found in natural scenes, an understanding must be generated of how sensitivity is affected by the size and location of stimuli within the visual field. This is of importance since we know that our ability to resolve details in an image decreases as the stimulation distance from the fovea increases (i.e., eccentricity). This has partially been attributed to a reduction in photoreceptor density as eccentricity increases and in a change of feature selectivity in the peripheral retina. Charactering the decrease in performance at detecting changes in the slope of the Fourier amplitude spectrum as a function of eccentricity will therefore generate a comprehensive understanding of how natural scenes are processed within the visual field as a whole.