HSWmark: Robust and reversible image watermarking with improved capacity using Quad-tree segmentation

This paper proposes a HSWmark (Histogram Shifting Watermark) based robust and reversible image watermarking scheme using Quad tree (QT) segmentation with improved capacity and minimum distortion. QT segmentation divides the input image into several sized non-overlapping blocks, and each block is used to embed payload data. Preprocessing is used to avoid overflow and underflow problems while embedding bits into pixels. Histograms of each block are generated and its maximum point and zero point are used for embedding. Payload embedding is performed using the Histogram Shifting (HS) method. Side information containing the details about QT segmentation structure, payload length, and compressed location map is stored in a few rows of cover images using Least Significant Bits replacement technique. At the extraction end, side information is extracted from the received image and payload data are retrieved back along with recovered cover images. Performance of the proposed scheme is assessed in terms of parameters Peak Signal to Noise Ratio (PSNR), Structural Similarity Index Measure (SSIM), Embedding Capacity (EC), Bit Rate and Bit Error Rate (BER). Parameter values are also compared with some existing techniques. Simulation results shows that attained PSNR is above 47 dB for all test images along with SSIM values higher than 0.979. Proposed technique is able to provide the maximum EC of 102,215 bits with 0.389 bit rate. Thus, the presented technique provides better visual quality of images with improved capacity. Proposed scheme is also tested for its robustness against different attacks and achieved PSNR is above 40 dB with SSIM value higher than 0.88 for all test images. Performance is also compared with the existing technique and from results it is clear that it provides better BER, PSNR and SSIM as compared to the existing techniques.