Region‐based scalable self‐recovery for salient‐object images

Abstract

Self‐recovery is a tamper‐detection and image recovery methods based on data hiding. It generates two types of data and embeds them into the original image: authentication data for tamper detection and reference data for image recovery. In this paper, a region‐based scalable self‐recovery (RSS) method is proposed for salient‐object images. As the images consist of two main regions, the region of interest (ROI) and the region of non‐interest (RONI), the proposed method is aimed at achieving higher reconstruction quality for the ROI. Moreover, tamper tolerability is improved by using scalable recovery. In the RSS method, separate reference data are generated for the ROI and RONI. Initially, two compressed bitstreams at different rates are generated using the embedded zero‐block coding source encoder. Subsequently, each bitstream is divided into several parts, which are protected through various redundancy rates, using the Reed‐Solomon channel encoder. The proposed method is tested on 10 000 salient‐object images from the MSRA database. The results show that the RSS method, compared to related methods, improves reconstruction quality and tamper tolerability by approximately 30% and 15%, respectively.

Keywords

• image authentication
• region‐based
• salient object
• scalable self‐recovery
• tamper detection