Research Article: The role of low-level image features in the affective categorization of rapidly presented scenes

Date Published: May 1, 2019

Publisher: Public Library of Science

Author(s): L. Jack Rhodes, Matthew Ríos, Jacob Williams, Gonzalo Quiñones, Prahalada K. Rao, Vladimir Miskovic, Sidney D’Mello.

http://doi.org/10.1371/journal.pone.0215975

Abstract

It remains unclear how the visual system is able to extract affective content from complex scenes even with extremely brief (< 100 millisecond) exposures. One possibility, suggested by findings in machine vision, is that low-level features such as unlocalized, two-dimensional (2-D) Fourier spectra can be diagnostic of scene content. To determine whether Fourier image amplitude carries any information about the affective quality of scenes, we first validated the existence of image category differences through a support vector machine (SVM) model that was able to discriminate our intact aversive and neutral images with ~ 70% accuracy using amplitude-only features as inputs. This model allowed us to confirm that scenes belonging to different affective categories could be mathematically distinguished on the basis of amplitude spectra alone. The next question is whether these same features are also exploited by the human visual system. Subsequently, we tested observers’ rapid classification of affective and neutral naturalistic scenes, presented briefly (~33.3 ms) and backward masked with synthetic textures. We tested categorization accuracy across three distinct experimental conditions, using: (i) original images, (ii) images having their amplitude spectra swapped within a single affective image category (e.g., an aversive image whose amplitude spectrum has been swapped with another aversive image) or (iii) images having their amplitude spectra swapped between affective categories (e.g., an aversive image containing the amplitude spectrum of a neutral image). Despite its discriminative potential, the human visual system does not seem to use Fourier amplitude differences as the chief strategy for affectively categorizing scenes at a glance. The contribution of image amplitude to affective categorization is largely dependent on interactions with the phase spectrum, although it is impossible to completely rule out a residual role for unlocalized 2-D amplitude measures.

Partial Text

Perceptual processing in the natural world is strongly influenced by motivational factors, allowing for adaptive behavioral routines in response to threats and opportunities in the environment [1–4]. Complex scenes can be affectively discriminated even with very rapid exposure times [5]. Enhanced brain physiological responses elicited by emotional, relative to neutral, scenes are detectable by ~200 ms using non-invasive recordings [6–8], although earlier latency modulations are apparent in intracranial studies [9–10]. Given how rapidly affectively salient information is extracted from these rich, visually cluttered stimuli, it remains to be understood how the human visual system accomplishes this feat [2].

We examined the utility of 2-D Fourier amplitude spectra in guiding the affective categorization of rapidly presented natural scenes. In a first step, we validated the existence of Fourier amplitude-based category differences through SVM classification based on amplitude-only features as inputs. The classification accuracy for pairwise comparisons of mutilation/neutral and disgust/neutral images was ascertained to be in the 70 to 75% range, well above chance. Subsequent findings from human observers who performed a rapid scene categorization task with backward masking indicated that AS information alone contributes only minimally to categorization performance. We observed a marked deterioration in categorization accuracy for both of the amplitude swapped image manipulations relative to intact scenes. Overall, these findings are consistent with evidence from animal classification tasks (e.g., [12, 42]) that low-level information provided by unlocalized 2-D Fourier amplitude spectra is not sufficient to enable high accuracy performance. To the extent that image amplitude contributes diagnostic information for categorizing scenes at a glance, it appears to be almost entirely dependent upon interactions with the image phase spectrum, which is generally viewed as conveying higher-level visual features (see also [69]).

 

Source:

http://doi.org/10.1371/journal.pone.0215975

 

Leave a Reply

Your email address will not be published.