The perception of naturalness correlates with low-level visual features of environmental scenes

Abstract

Previous research has shown that interacting with natural environments vs. more urban or built environments can have salubrious psychological effects, such as improvements in attention and memory. Even viewing pictures of nature vs. pictures of built environments can produce similar effects. A major question is: What is it about natural environments that produces these benefits? Problematically, there are many differing qualities between natural and urban environments, making it difficult to narrow down the dimensions of nature that may lead to these benefits. In this study, we set out to uncover visual features that related to individuals' perceptions of naturalness in images. We quantified naturalness in two ways: first, implicitly using a multidimensional scaling analysis and second, explicitly with direct naturalness ratings. Features that seemed most related to perceptions of naturalness were related to the density of contrast changes in the scene, the density of straight lines in the scene, the average color saturation in the scene and the average hue diversity in the scene. We then trained a machine-learning algorithm to predict whether a scene was perceived as being natural or not based on these low-level visual features and we could do so with 81% accuracy. As such we were able to reliably predict subjective perceptions of naturalness with objective low-level visual features. Our results can be used in future studies to determine if these features, which are related to naturalness, may also lead to the benefits attained from interacting with nature.

Figure 1. Scree plots for the First set (left) and the Second set (right) of images.

Stress values are plotted as a function of the dimensionality in which the MDS data were scaled.

Figure 2. Plotted results of MDS dimensions 1 (X-axis) and 2 (Y-axis) for the first set, with pictures superimposed.

The pictures are placed in the image based on their weights on dimension 1 and 2. A subset of the 70 images is plotted here because there are too many images to make this plot readable.

Figure 3. Plotted results of MDS dimensions 1 (X-axis) and 2 (Y-axis) for the second set, with pictures superimposed.

The pictures are placed in the image based on their weights on dimension 1 and 2. A subset of the 70 images is plotted here because there are too many images to make this plot readable.

Figure 4. Correlation of Perceived Naturalness with weights on Dimension 1 for the first set and second sets of images.

Figure 5

a) A sample image (b) Image's saturation map (c) Image's hue map (d) Image's brightness map.

Figure 6. Comparison of two images in their color diversity properties.

a) SDHue  = 0.11, SDSat = 0.22, SDbright = 0.21 b) SDHue = 0.19, SDSat = 0.26, SDbright = 0.26.

Figure 7. Comparison of two images in their Entropy a) Entropy = 7.63 b) Entropy = 7.10.

Figure 8

(a) Sample image, (b) the detected straight edges of the sample image(c) the edge density map of the sample image.

Figure 9. Correlation Matrix of low-level visual features with perceived naturalness.

The color bar indicates the strength of the correlation from -.8 to 1.

Figure 10. Feature weights for the LD classification algorithm in predicting high- vs. low- perceived naturalness of the images.

A high absolute value of the weight indicates that that feature is important for classification. A positive weight indicates that that increasing this feature would lead to increased perceived naturalness; a negative weight indicates that increasing this feature would lead to a decrease in perceived naturalness. Error bars reflect 2 standard deviations from the mean.

Figure 11. Correlation Matrix of low-level visual features with the first 4 Principal Components across the 307 images.

The color bar indicates the strength of the correlation from -.4 to +.4