ICS Colloquium – Dr. Aaron Johnson

• In-person event

Aaron Johnson Ph.D. Assistant Professor, Department of Psychology,
Concordia University.

Title: Human perception of natural scenes.

The environment that humans inhabit, and the tasks that we must perform have shaped the design and function of our visual system through evolution and experience. Consequently, a research field has recently emerged that focuses on natural images (i.e. real-world stimuli as opposed to synthetic stimuli). When analyzing single objects, it is possible to describe objects as having a small number of visual features (e.g., orientation, luminance, colour). This type of analysis has been used with great success in the past; yet real world objects are rarely isolated. In a natural image with many objects, it is not enough to give descriptions of all the objects separately. First, more global descriptors have to be defined and extracted. To describe these, researchers have started to use statistical methods, i.e., characterize the whole image by distributions of visual features, the characteristics of these distributions, and their variations across space. In my talk, I will describe my work on discovering the existence of characteristic information within image statistics, and show how they differentiate between different types of scenes. Further, I will show evidence that humans are sensitive to these image statistics, and their variations. Finally, I will show how image statistics might be used in daily visual tasks such as texture segmentation and scene recognition.