Reversible data hiding in encrypted DICOM images with fixed and block-wise pixel prediction

Abstract

Reversible Data Hiding in Encrypted Images (RDHEI) is a technique that enables additional data to be embedded into encrypted images while preserving the ability to fully recover both the original image and the hidden information, making it particularly valuable for applications requiring confidentiality and integrity, such as medical imaging. This paper presents a high-capacity reversible data hiding scheme for encrypted DICOM images, addressing the unique challenges posed by their 16-bit pixel depth and structured entropy distribution. The proposed method introduces a binary decomposition strategy that separates the image into two complementary components, enabling tailored prediction techniques for each part. The first component is processed using fixed prediction—a lightweight bit-flipping mechanism, while the second employs variable block-wise model-based prediction optimized for low-error encoding. To reduce the auxiliary data overhead introduced by this two-phase preprocessing, two compression strategies—Huffman coding and Extended Run-Length Encoding—are employed. Experimental results on anonymized DICOM datasets show that the method achieves embedding rates exceeding 10 bpp while maintaining full reversibility. Comparative analysis confirms the method’s competitiveness with recent state-of-the-art RDHEI schemes. The approach is also benchmarked on non-DICOM datasets to demonstrate general applicability.