Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Structuring Quantitative Image Analysis with Object Prominence

Christian Arnold, Andreas Küpfer

TL;DR

This article exemplifies object prominence with different implementations -- object size and centeredness, the pixels' image depth, and salient image regions -- and showcases the usefulness of the approach with two applications.

Abstract

When photographers and other editors of image material produce an image, they make a statement about what matters by situating some objects in the foreground and others in the background. While this prominence of objects is a key analytical category to qualitative scholars, recent quantitative approaches to automated image analysis have not yet made this important distinction but treat all areas of an image similarly. We suggest carefully considering objects' prominence as an essential step in analyzing images as data. Its modeling requires defining an object and operationalizing and measuring how much attention a human eye would pay. Our approach combines qualitative analyses with the scalability of quantitative approaches. Exemplifying object prominence with different implementations -- object size and centeredness, the pixels' image depth, and salient image regions -- we showcase the usefulness of our approach with two applications. First, we scale the ideology of eight US newspapers based on images. Second, we analyze the prominence of women in the campaign videos of the U.S. presidential races in 2016 and 2020. We hope that our article helps all keen to study image data in a conceptually meaningful way at scale.

Structuring Quantitative Image Analysis with Object Prominence

TL;DR

This article exemplifies object prominence with different implementations -- object size and centeredness, the pixels' image depth, and salient image regions -- and showcases the usefulness of the approach with two applications.

Abstract

When photographers and other editors of image material produce an image, they make a statement about what matters by situating some objects in the foreground and others in the background. While this prominence of objects is a key analytical category to qualitative scholars, recent quantitative approaches to automated image analysis have not yet made this important distinction but treat all areas of an image similarly. We suggest carefully considering objects' prominence as an essential step in analyzing images as data. Its modeling requires defining an object and operationalizing and measuring how much attention a human eye would pay. Our approach combines qualitative analyses with the scalability of quantitative approaches. Exemplifying object prominence with different implementations -- object size and centeredness, the pixels' image depth, and salient image regions -- we showcase the usefulness of our approach with two applications. First, we scale the ideology of eight US newspapers based on images. Second, we analyze the prominence of women in the campaign videos of the U.S. presidential races in 2016 and 2020. We hope that our article helps all keen to study image data in a conceptually meaningful way at scale.
Paper Structure (16 sections, 2 equations, 4 figures, 1 table)

This paper contains 16 sections, 2 equations, 4 figures, 1 table.

Table of Contents

  1. Introduction
  2. Object Prominence as Structuring Principle for Image Analysis
  3. Analyzing Images with Semantic Units
  4. Semantic Units that Matter
  5. Measuring Object Prominence
  6. Identifying Semantic Units
  7. Mapping the Salience of Semantic Units
  8. Size and Centeredness
  9. Depth
  10. Salient Object Detection (SOD)
  11. Applications
  12. Improving Idealpoint Estimation from Visual Bag of Words (VBoW)
  13. When Importance itself is a Relevant Object of Study: Gender Bias in US Presidential Campaign Videos
  14. Conclusion
  15. Application 1: Idealpoint Estimates
  16. ...and 1 more sections

Figures (4)

  • Figure 1: Example images from our corpus and their salient regions detected by MBD Zhang2015minimum. A brighter color indicates higher salience.
  • Figure 2: Idealpoint estimates for Climate Change in four different scenarios based on images or text. Ground truth is based on a random sample of 120 news article texts (15 per news outlet); All other models use the full population of 2853 news articles.
  • Figure 3: Coefficients of interaction effect for two models. Face Depth Position has the normalized depth of a face as the outcome; Face Size represents the relative size of a face. The models depend on 67,782 recognized faces and include fixed effects for candidate ID, election year, and the candidate's visibility in the video. The errors are clustered on individual video frames.
  • Figure 4: Full overview of idealpoint estimates in all scenarios across five issues and eight news outlets.