HoSVD-NSST Framework for Secure Dual Medical Image Watermarking Using Deep Learning

Roopkamal K. Thind, Rajat Sood, Jyoti Rani, Ashima Anand, Anamika Sharma

Abstract: The rapid rise in the amount of medical data poses a significant challenge for researchers in preventing its manipulation. To overcome this challenge, this research introduces a modern watermarking technique for concealing two secret images within a cover image, employing Higher Order Singular Value Decomposition (HOSVD), Non-Subsampled Shearlet Transform (NSST), and a 12-layer deep neural network architecture. The proposed methodology initially decomposes the cover image with HOSVD to acquire factor matrices and a core tensor. The core tensor is subjected to NSST transformation, producing many components, with the low-frequency component particularly chosen for embedding the hidden images. A complex 12-layer neural network architecture is utilised to effectively embed two secret images into the low-frequency component while preserving the visual integrity of the cover image. The extraction procedure reflects this architecture with a reveal network capable of effectively recovering both concealed images. Experimental findings illustrate the efficacy of the proposed strategy regarding imperceptibility, resilience against prevalent attacks, and enhanced capacity relative to existing techniques.

External IDs:dblp:conf/iciap/ThindSRAS25