DEVELOPMENT OF HIGH-CAPACITY REVERSIBLE DATA HIDING METHODS FOR ENCRYPTED IMAGES

Abstract

Security issues in image handling have gained significant importance because of the increasing use of outsourced storage facilities for handling images through operations such as storage, transmission, indexing, and manipulation. Confidentiality and data association are two aspects that should be taken into account when dealing with images in such an environment since one aspect of confidentiality is that the information embedded within the images should not be disclosed, and at the same time, auxiliary information, such as authentication, ownership, indexing, or medical data, should also be associated with the images. In view of these requirements, RDH EI technology has been developed that enables the embedding of auxiliary information in encrypted images with the ability to reconstruct the original image without any error. This work considers two different RDH-EI approaches. The first method is related to room reservation prior to encryption in order to keep the embedding space intact, whereas the second approach relates to embedding a higher amount of information after encryption using block index modulation. The first contribution develops an Adaptive Weighted Predictor for a checkerboard based RDH-EI framework. The image is separated into dot and cross pixel sets. Dot pixels are predicted first with a Rhombus Mean Predictor, and the error map obtained from this stage is reused as a local reliability indicator for cross-pixel prediction. Instead of treating all contextual predictors equally, the proposed predictor assigns lower influence to neighborhoods with larger first-stage prediction errors and higher influence to more reliable neighborhoods. A damping parameter is incorporated to avoid excessive compensation in smooth regions. The reported evaluation demonstrates improvement in MAE, MSE, and prediction-error variance with respect to the RMP plus AMP baseline. The second contribution introduces an index-modulated high-capacity RDH-EI method for distortion-free data embedding in encrypted images. The method employs dual-stage encryption, combining block permutation with intra-block pixel scrambling to protect visual content while enabling structured encrypted-domain embedding. Through adaptive block regulation, encrypted blocks are arranged in a v deterministic order, and secret bits are represented by selecting valid permutation indices rather than by directly modifying encrypted pixel values. This index-based embedding strategy preserves reversibility, allowing complete recovery of both the embedded auxiliary information and the original image. Experimental results demonstrate stable high-capacity embedding performance across standard benchmark images, while perfect reconstruction is consistently achieved for the recovered images. Together, the two contributions investigate separate but related bottlenecks in RDH EI design. The first improves the compactness of prediction errors in an RRBE oriented workflow, whereas the second increases encrypted-domain embedding capability within a VRAE-style setting.