Estimating Color-Concept Associations from Image Statistics

Abstract

50 0 50 100 b* R O Y G Br Gr Pu Pi Pi B r = 1 Balls Sectors Categories Δr = 1 Δr = 1 Δh = 5°Δ r = 1 Δh = 40°Δ r = 20 Δh = 5°Δ r = 20 50 0 -50 -50 Δr = 20 Δh = 40°a * Figure 1. We constructed models that estimate human color-concept associations using color distributions extracted from images of relevant concepts. We compared methods for extracting color distributions by defining different kinds of color tolerance regions (white outlines) around each target color (regularly spaced large dots) in CIELAB space. Subplots show a planar view of CIELAB space at L* = 50, with color tolerance regions defined as balls (left column; radius ∆r ), cylindrical sectors (middle columns; radius ∆r and hue angle ∆h), and category boundaries around each target color (right column; Red, Orange, Yellow, Green, Blue, Purple, Pink, Brown, Gray; white and black not shown). Each target color is counted as "present" in the image each time any color in its tolerance region is observed. This has a smoothing effect, which enables the inclusion of colors that are not present in the image but similar to colors that are. A model that includes two sector features and a category feature best approximated human color-concept associations for unseen concepts and images (see text for details).Abstract-To interpret the meanings of colors in visualizations of categorical information, people must determine how distinct colors correspond to different concepts. This process is easier when assignments between colors and concepts in visualizations match people's expectations, making color palettes semantically interpretable. Efforts have been underway to optimize color palette design for semantic interpretablity, but this requires having good estimates of human color-concept associations. Obtaining these data from humans is costly, which motivates the need for automated methods. We developed and evaluated a new method for automatically estimating color-concept associations in a way that strongly correlates with human ratings. Building on prior studies using Google Images, our approach operates directly on Google Image search results without the need for humans in the loop. Specifically, we evaluated several methods for extracting raw pixel content of the images in order to best estimate color-concept associations obtained from human ratings. The most effective method extracted colors using a combination of cylindrical sectors and color categories in color space. We demonstrate that our approach can accurately estimate average human color-concept associations for different fruits using only a small set of images. The approach also generalizes moderately well to more complicated recycling-related concepts of objects that can appear in any color.

Semantic Discriminability for Visual Communication

“…However, interpretability can also be achieved when not all color-concept pairs are semantically resonant [31], see Section 2.2. Rathore et al [28] referred to this more general case as semantically interpretable color palettes. For simplicity, we use the term interpretability in the present work to refer to the more general case.…”

“…We used a previous dataset on color-concept associations from Rathore et al [28] to select the colors for the present study (Section 3.1), define accuracy for the present tasks (Section 3.2.1), and quantify semantic distance (Section 3.2.2). In [28], participants rated association strengths between each of 12 fruits and each of 58 colors (UW-58 colors), uniformly sampled in CIELAB space (∆E = 25). This distance should be at least one noticeable difference [34,35].…”

“…This distance should be at least one noticeable difference [34,35]. Further details on the methods of [28] are in Supplementary Material of the present paper. 3), with the size of the marks corresponding to association strength with the concept named at the top of the column (same data as in A.1).…”

“…We chose the colors and fruit concepts using the mean color-concept association data from Rathore et al [28] (see Table S.6 in the Supplementary Material) and color distances in CIELAB space (∆E). In Experiment 1, we selected eight colors and two fruits to have the following properties.…”

Shared understanding of color among congenitally blind and sighted adults

“…However, these prior studies may underestimate the capacity of language to transmit color information. Like most studies of color knowledge in sighted people, these studies focused on knowledge of associative color facts such as that strawberries are red, rather than on inferentially rich, causal understanding of color (e.g., [39][40][41][42][43]. Such color factoids might be least likely to be culturally transmitted since, for both sighted and blind people alike, they are inferentially shallow and disconnected from other things we know about objects.…”