Bitwise level image permutation using a sinusoidal signal

Abstract

This paper presents a permutation technique based on a sinusoid whose complexity is O(N) but providing a very acceptable security. It considers an image as a set of vertical and horizontal bits that shift under the effect of a sinusoidal wave. This results in the permutation of the image content at the level of bits. As a result, for only one round of the algorithm, the correlation between pixel values may reach 0.03. Moreover, the proposed method offers the opportunity to permute a selective free form. The algorithm is able to consider any concave or convex curve in the selected part. The significance of these two criterion is in the possibility to mix the technique together with an object detection application in order to encrypt the object of interest in a live broadcast.