Automatic Classifier Based Reversible Image Transformation for Reversible Data Hiding in Encrypted Images

Abstract

Reversible data hiding in encrypted image is one of the method to protect data in cloud network. This article proposes data hiding algorithm that use reversible data hiding in encrypted images (RDH-EI) and reversible image transformation. It is easy to understand that the encrypted image in cipher format contains some information. But in reversible image transformation, the content of original image is transformed to another transformed image that looks similar to target image. Usually a particular percentage of blocks are classified as two classes for image transformation. In proposed method a classifier is introduced that classifies the pixels into two classes using Otsu's threshold for image transformation. This transformed image acts as the encrypted image. This transformed image is uploaded to the cloud. In the cloud network, data can be embedded to the encrypted image (transformed image) using a reversible data hiding approach. Therefore the data embedded at the cloud network does not depend on the reversible image transformation at the encoder. Similarly the data extraction process does not depend on the inverse reversible image transformation at the decoder. The performance of the proposed reversible image transformation is measured by using metrics like MSE, PSNR and Correlation Coefficient. The performance and quality shows that proposed system is better than the conventional reversible data hiding method that uses reversible image transformation.