On measuring the local perceived blur in natural scenes both in the lab and online
Natural retinal images contain blur. Image blur varies across the visual field depending on the three dimensional image structure and the observers’ point of accommodation. There is currently no consensus in the field on how to quantify local perceived blur in natural images. We developed a biologically-plausible measure of local image blur, local slope, which is computed from the output of a set spatial frequency and orientation band-limited filters. To determine if local slope captured the perception of image blur variation, we used a task in which observers traced regions in natural images that matched the perceived blur of a standard blur. We ran our tracing task both under controlled laboratory conditionswith a small number of observers and online using a large sample of Amazon Mechanical Turk Workers. There was good agreement across observes on image regions with matched perceived blur and a correlation between local slopes of blur-matching areas of natural and standard images for data collected in the lab and online. These data indicate that local slope may provide a rapid and simple metric of perceived blur in natural scenes for human observers.