To See or Not To See: Determining the Recognition Threshold of Encrypted Images

Abstract

There are numerous standards and recommendations when it comes to the acquisition of visual quality assessment from human observers. The recommendations deal with clearly visible images and try to keep the just-noticeable-difference between quality steps as small as possible to facilitate an exact measurement of image differences. When it comes to the assessment of selective encryption schemes the question is the opposite. The quality is not really of interest, the question is rather if the content of the images is discernible at all. There are no recommendations in literature for this kind of task. In this paper we will outline different protocols and setups, test them and form a recommendation for the acquisition of the recognition threshold for encrypted images from human observers.