Difference expansion based reversible data embedding and edge detection

Abstract

Difference Expansion (DE) based techniques are evolving in the last decade. The DE based techniques aims at reversibly embedding the data into cover content. Most of the DE based reversible embedding techniques aims to reduce/avoid the amount of auxiliary information to be embedded into the cover images and then state that the auxiliary information generated in the process has no applications rather to keep track of the overflow and/or underflow problems that may arise during the reversible embedding. Our objective is to address the above issues as well as address other related issues as stated below. In this article, we present multiple contributions based on difference expansion technique. At first, we propose a DE based reversible scheme to reduce the amount of auxiliary information to be embedded into the image, to increase the length of the watermark, using the pixel pair information of the changeable differences. The application which emphasizes the need for preserving the auxiliary information, generated by layer-2 DE embedding, is identified. A simple mechanism is suggested to increase the visual quality when the data is embedded using layer-2 embedding. Next, we propose an edge detection scheme using the DE technique. We used PSNR and SSIM visual quality assessment metrics and embedding capacity measure to assess the effectiveness of the proposed data embedding algorithms. The metrics Precision and Recall are used to assess the DE based edge detection scheme. The experimental tests have been conducted on two different datasets having complex visual distinctness features. The analytical and experimental results demonstrate that the proposed work significantly improved state of the art.