Deep learning‐based robust medical image watermarking exploiting DCT and Harris hawks optimization

Abstract

Image watermarking is an effective way to secure the ownership of digital photographs. This paper proposes a new methodology for integrating a watermark on the basis of various integrative strengths. The image is separated as 8 × 8 pixels blocks that do not overlap. The pixel size for each image block has been determined. For the embedding areas, picture blocks with the highest value have been chosen. Therefore, discrete cosine transformation (DCT) is transformed. The DCT coefficients are chosen in the midfrequency and the average selected DCT blocks are determined using a series of rules to produce various integration strengths. The watermarking bits were merged with the proposed deep learning convolution neural network (DLCNN) through a series of integration standards. The binary watermark has been scrambled by an Arnold transform until it is incorporated for additional stability. During the image carrier, a pattern recognition model depending on DLCNN is utilized to identify and extract the watermark and to recognize the watermark using the Harris hawks optimization (HHO) algorithm. The findings of the tests demonstrated that the system suggested is most imperceptible than the other current systems. The proposed method attains the efficiency watermarked picture with 46 dB peak signal-to-noise ratio value. This paper focuses on robust medical image watermarking exploiting DCT by using the HHO algorithm. The watermark lossless compression reduces watermark payload without data loss. In this research work, watermark is the consolidation of DCT and image watermarking secret key. The performance of robust medical image watermarking exploiting DCT with the HHO algorithm is compared with other conventional compression methods. HHO is found better and used to control watermarked image degradation in medical images watermarking. The proposed system also created a high resistance to remove watermarks during many attacks.