A Novel Reversible Ternary Embedding Algorithm Based On Modified Full Context Prediction Errors

We propose a high capacity reversible ternary embedding-watermarking algorithm based on a modification of full-context-prediction-errors (MFCPE) wherein the binary bit stream is converted to the ternary stream then error histogram shifting is utilized to embed the ternary stream. Unlike the existing predictor methods, we provide a full context prediction with a modification of each pixel at most by 1, which significantly reduces distortion. Experimental results confirm that the proposed algorithm achieves high PSNR while providing a higher embedding capacity. Also, results indicate that MFCPE outperforms the existing methods in terms of payload and the watermarked image quality.