Watermarking scheme using slantlet transform and enhanced knight tour algorithm for medical images

Abstract

Digital watermarking has been employed as an alternative solution to protect the medical healthcare system with a layer of protection applied directly on top of data stored. Medical image that is highly sensitive to the image processing and cannot tolerate any visual degradation has become the focus of digital watermarking. However, since watermarking introduces some changes on medical images, it is a challenge for medical image watermarking to maintain high imperceptibility and robustness at the same time. Research to date has tended to focus on the embedding method instead of the sequence of embedding of the watermarking itself. Also, although watermarking has been introduced into medical images as a layer of protection, it still cannot prevent a knowledgeable hacker from retrieving the watermark. Therefore, this research proposes a robust watermarking scheme with high imperceptibility for medical images to increase the effectiveness of the medical healthcare system in terms of perceptibility, embedding technique, embedding region and embedding sequence of the watermarking scheme. To increase imperceptibility of a watermark, this research introduces Dynamic Visibility Threshold, a new parameter that increases visual quality in terms of imperceptibility. It is a unique number which differs for each host image using descriptive statistics. In addition, two new concepts of embedding region, namely Embeddable zone (EBD) and Non-Embeddable zone (NEBD) to function as a non-parametric decision region to complicate the estimate of the detection function are also proposed. The sequence of embedding is shuffled using enhanced Knight Tour algorithm based on Slantlet Transform to increase the complexity of the watermarking scheme. A significant result from the Peak Signal-to-Noise Ratio (PSNR) evaluation showing approximately 270 dB was obtained, suggesting that this proposed medical image watermarking technique outperforms other contemporary techniques in the same working domain. Based on the experimental result using the standard dataset, all host images are resilient to Salt and Pepper Noise, Speckle Noise, Poisson Noise, Rotation and Sharpen Filter with minimum Bit Error Rate (BER) of 0.0426 and Normalized Cross-Correlation (NCC) value of as high as 1. Since quartile theory is used, this experiment has shown that among all three quartiles, the Third Quartile performs the best in functioning as Dynamic Visibility Threshold (DVT) with 0 for BER and 1 for NCC evaluation.