What Is a Clear Image? Human Sensitivity to Noise in Naturalistic Images

Our daily lives are full of visual noise: from smog to fog to static on a television. Even with noise obscuring our view, we are able to perceive the world and judge visual quality in real-time. How do we judge image quality, and what is it about an image that clues us into its quality? We set out to determine what features of images we use to judge quality in an image by presenting images (aerial shots, landscapes, and medical images) with noise of varying level added to obscure either global or local features of an image to naïve observers. We analyzed the images for different visual features and compared human decisions with classification by a binary tree using said analyzed features. We demonstrate that human judgment of image quality is dependent on the perception of local image features, not on global features, image class or noise level.