Attentive texture similarity as a categorization task: Comparing texture synthesis models

Abstract

Many attempts have been made to characterize latent structures in "texture spaces" defined by attentive similarity judgments. While an optimal description of perceptual texture space remains elusive, we suggest that the similarity judgments gained from these procedures provide a useful standard for relating image statistics to high-level similarity. In the present experiment, we ask subjects to group natural textures into visually similar clusters. We also represent each image using the features employed by three different parametric texture synthesis models. Given the cluster labels for our textures, we use linear discriminant analysis to predict cluster membership. We compare each model's assignments to human data for both positive and contrast-negated textures, and evaluate relative model performance.

Fig. 1

Examples of synthetic textures obtained from the texture synthesis models under consideration in this experiment. The original target texture is pictured at left. There is a striking increase in quality across the different models.

Fig. 2

The 30 “training” images drawn from the Brodatz database. Subjects placed these images into clusters according to visual similarity in an unconstrained sorting task.

Fig. 3

(Lower right) A Scree plot of encoding error vs. number of clusters for positive textures. A seven-cluster solution was selected on the basis of this plot, and the texture clusters thus selected are also pictured with brief high-level descriptors of each group (left).

Fig. 4

(Lower right) A Scree plot of encoding error vs. number of clusters for negative textures. A seven-cluster solution was selected on the basis of this plot, and the texture clusters thus selected are also pictured with brief high-level descriptors of each group (left).

Fig. 5

Plots of leave-one-out accuracy as a function of the number of principal components used to summarize the features contained in our three models. The left column displays the results obtained from classification of the positive textures and the right column shows the results obtained from negative textures. Our three discriminant functions (described in the text) are shown by rows with the ‘diagonal linear’ results at the top, ‘linear’ results in the middle, and ‘diagonal quadratic’ results at the bottom.