In order to develop a method for mining facial photographs for hidden information, psychologists Rob Jenkins and Christie Kerr performed a study in which they captured high-resolution images of subjects’ faces and used image analysis tools to zoom in on the subjects’ faces and extract images of unseen bystanders from reflections in the eyes.
Jenkins (University of York) and Kerr (University of Glasgow) used eight volunteer photographic subjects for their research. Each volunteer served as the subject of a single photograph and as a bystander in three other photographs. A Hasselblad H2D 39 MPixel camera with 120 mm macro lens was used to photograph the subjects from a distance of 1m away. This camera, which is now discontinued, features a CCD image sensor and an electronically-controlled integrated shutter. With 5,412 x 7,216 pixel images, the researchers were able to identify the volunteers in the corneal reflection image of three different subjects. For purposes of the study, the largest reflection of each bystander’s face was selected for presentation for a face matching experiment.
In the experiment, images of 27 to 36 pixels wide by 42 to 56 pixels high were selected for the experiment. For presentation in the experiment, the extracted face images were rescaled to a height of 400 pixels. Brightness and contrast were automatically adjusted using the auto contrast function in Adobe Photoshop to improve image definition. In order to determine whether the corneal reflection images could be used for face matching, the researchers paired each of these images with a standard photo of the same face or a similar-looking face and observers were asked to make a judgment.
Page 1 | Page 2
