Prediction error expansion based reversible watermarking scheme in higher order pixels

Abstract

In today’s digitally-driven world flooded with digital content and the widespread availability of images online, digital image watermarking is crucial. It ensures copyright protection, maintains content authenticity, and preserves the integrity of visual information. In this paper, a reversible watermarking technique is proposed based on splitting the gray scale image pixel’s intensity value and conventional prediction error expansion. To begin with, the pixel values of an image are split and posted into two distinct groups. The values in the hundred’s and ten’s place are categorized as the higher order pixel group (\(\alpha\)), while the values in the units place are categorized as lower order pixel group (\(\beta\)). Through this splitting process, it has been noticed that numerous adjacent pixels share identical values. In order to embed data, the conventional Prediction Error Expansion technique is employed specifically for the higher order pixel group. After determining the prediction error, it will be expanded and secret data is embedded into it. This scheme has been experimentally verified that it is superior than few existing state of art schemes. With a high embedding capacity of about 65335 bits in a single 512 \(\times\) 512 image, the watermarked images exhibited a PSNR of approximately 36.34 dB and an SSIM of around 0.94. Payloads for 10,000 and 20,000 bits are tested and their PSNR values are 44.56 dB and 41.35 dB respectively.