INTERPOLATION BASED ADAPTIVE REVERSIBLE DATA HIDING FOR DIGITAL IMAGE APPLICATIONS

Abstract

Reversible Data Hiding (RDH) has emerged to be a significant multi-disciplinary research area. Different types of RDH schemes have already been developed and demonstrated their potentials for applications like media-content authentication, copy-right protection, integrity establishment and annotation. Recently, interpolation based RDH (IRDH) schemes have shown great promises for better rate-distortion performance with a set of unique advantages over their counterparts. Despite great promises, existing IRDH schemes, however, have no consideration of ‘effective’ capacity management for varying size of the payload. As a result, disregarding the varying embeddingcapacity requirement, existing IRDH schemes fail to maintain the minimum possible distortion in the embedded image. An IRDH scheme with adaptive embedding is therefore vital for better rate-distortion performance. In this thesis, a new IRDH scheme is developed and presented based on a simplified parabolic interpolation (SPI) and adaptive embedding. Additionally, an adaptive embedding approach is introduced with two new techniques, where a capacity control parameter is formulated and utilized to attain the varying capacity requirement by determining a minimum set of embeddable bits in a pixel for the best possible embedded image quality. The first technique introduces a flag bit for distinguishing the original payload bits from their complement which are embedded in the interpolated pixels. In the second embedding technique, the correlation between an embeddable pixel and estimated versions of an embedded pixel is utilized to avoid the use of any flag bit as such an extra payload bit per pixel can be embedded therein. Thereby, the embedding capacity significantly increases without compromising the image quality for similar values of the capacity control parameter. Computational modeling of the new IRDH scheme is presented and its performance is evaluated with a set of popular test-images. While compared with the prominent IRDH schemes, the proposed scheme both with or without flag demonstrated its efficiency for significantly better embedding rate-distortion performance. Moreover, being up-sampled, the embedded image would have a higher spatial resolution. It also does not require any location map, and thus the total capacity can be effectively used for data embedding. Nevertheless, it keeps the original pixels untouched and thus, would be useful in military and medical image applications that restrict minimum possible changes in the cover images.