A structural similarity assessment for generating hybrid images

We discuss a structural similarity evaluation method in order to generate hybrid images for graphical password. The hybrid image consists of two images. One is edge information of a key image and the other is coarse features of a decoy image. Visibility of the hybrid image depends on the viewing distance and the structural similarity of the images. This property of the hybrid image is useful for preventing shoulder-surfing of graphical password. However, an evaluation method for checking whether the decoy image is appropriate to overlay the key image has not been established. To address this issue, we utilize Speeded Up Robust Features (SURF) to evaluate structural similarity between two images. Simulation results show that our similarity measure mostly coincides with the visibility of a key image. We also compare our objective metric with the structural similarity (SSIM) index and subjective scores obtained from a user test based on Scheffe's method of paired comparisons.