Item

Abstract

Besides color and shape, texture is one of the basic dimensions of images. While there
has been some research on the psychological dimensions of gray textured patterns, up
to now, there has been reported only one study on the categories of human similarity
ratings of color patterns (Mojsilovic et al., 2000, IEEE Transactions on image processing).
To test the reliability of the reported relevant pattern features, we repeated the
study with new stimuli (20 colored patterns, which were chosen in a pre-test from a
large database of patterns). These were presented on a large computer display (24")
in 40 rounds with one target pattern and 10 other test patterns at a time. The participants
had to indicate the perceived similarity of the test patterns by placing them
in an accordant horizontal distance to the target pattern. The distance was encoded
as dissimilarity measurement on a scale from 1 (very similar) to 100 (totally dierent).
Using multidimensional scaling techniques, four independent dimensions could be extracted
from the participants' ratings: 1. Directionality. This dimension differentiates
between the patterns with continuous lines and the patterns with closed and rounded
lines. 2. Color purity. This dimension dierentiates between the black-and-white and
pale patterns and the patterns with saturated and vivid colors. 3. Color tone. In the
overall-solution, this dimension differentiates between patterns containing red and the
other patterns. Additional single-case-analyses indicated that \red" as a distinctive
feature of pattern similarity was a \majority decision"; at least some participants favored
\green" as being distinctive against all other colors. 4. Regularity, Orientation
and Complexity. Basically, the structural blueprints of the patterns in this dimension
change from being symmetrical and orthogonal on one side over being diagonally oriented
and of disturbed symmetry to complete asymmetry on the other side. The four
dimensions account for 54% of the overall variance of the participants' ratings. Several
psychologically relevant features from the foregoing study could be replicated (directionality,
regularity, orientation, complexity and color purity), indicating their general
importance in pattern similarity judgments. Two other dimensions did not play a
role in our outcomes (\pattern heaviness" and chromaticity), and our dimension \color
tone" did not show up in the earlier study. Seemingly, the saliency of these features
depends more on the stimulus material than the foregoing.