Scientists from the University of Glasgow and the University of York (United Kingdom) have published a work describing the technique of searching for “hidden” faces in photographs by reflections in the pupil. The authors of the study believe that this method can be used for forensic examination. For example, in the pupils in the photograph of the victim may be the face of the offender. Frames from a camera seized during the investigation can also be studied to restore the picture of what happened. Actually, the method of identifying "hidden" faces is quite banal: the photo is simply zoomed in a certain place, then the contrast increases and a special software is launched to increase the sharpness. Sharpening algorithms have recently improved markedly and show impressive results. For example, this is what SmartDeblur is capable of by Russian developer Vladimir Yuzhikov.
Source image ')
The result of processing.
The same technique can be used to increase the sharpness on blurred images of faces. The illustration shows the faces found in the pupils in the photographs. The authors of the scientific work established that the participants in the experiment recognize such persons with an accuracy of 90%, which is quite suitable for forensic examination.
The size of the face in the pupil is about 30,000 times smaller than the face of the subject. Accordingly, you can roughly understand what size the original image should be in order to restore the reflection. As the research has shown, a picture of 320 pixels is enough to recognize a person. For example, the photo below has a size of 16x20 pixels.
Thus, the size of the original image should be about 9.6 megapixels, if there is only a face in the frame. Recently, even in social networks you can find a lot of high-resolution photos. Users publish them and do not suspect that they may accidentally give out some kind of secret. The quality of photographic equipment is getting better, so in the future it will be necessary to take into account that the pupils work like a mirror.
The authors of the research process processed high-quality 39-megapixel photos taken by a professional Hasselblad camera. It will take several years - and the same resolution will be on the cameras in smartphones.