Analysis of the use of some statistical measures in deciding about the efficiency of an image encryption algorithm

Most of papers on image encryption provide some measures to attest that the presented works satisfy the required criteria in the field. Some others propose a comparative study and decide on whether one work is better than another one always using these measures. These last are of different types where the correlation coefficient is one among them. It is a statistical measure used to analyze the resemblance between neighboring pixels in an obtained cipher. In this paper, we give a study about the relevance of this criterion to the decision about the quality of an algorithm. We demonstrate that its importance is limited to have an idea about the resulting decorrelation. However, its use to decide about the quality of the algorithm, or to compare it to another one is not possible. We show that any calculated value is surrounded by randomness.