Privacy preserving scheme for online image sharing

Abstract EN

This paper presents a privacy protection solution for online photo share by obscuring faces in images, which keeps persons anonymous.

The proposed system contains two modules.

The first is the face detection module to identify region of interest (ROI) which is faces.

The second is a face encryption module, which encrypts the ROI pixels using keys so that the access to the faces is restricted.

The face detection module uses skin color detector in YCbCr color space to detect skin areas in the image.

Also, to overcome the illumination problems in color images, two color constancy methods were adopted for color correction and lighting of the input images.

The edges of the image are utilized to separate the faces segments from the background or object that have similar skin color.

Morphological operations such as erosion are applied to remove small areas and hole filling to remove any holes in the binary segments.

In addition, the correct faces are located by using a set of features.

The face encryption module uses two chaotic logistic maps.

One map is used for shuffling the face area pixels and another map is used for encrypting the pixels.

Both shuffling and encryption are done using a keys.

The face detection was tested on Caltech face database and showed a high detection rate and can localize face under different illumination conditions.

The experiments on face encryption showed satisfactory results in various tests in terms of key space, PSNR, MSE and entropy analysis.