Watermarking of Frequency and Steganography for Protection of Medical Images Based on Bacterial Foraging Optimization and Genetic Algorithm

Abstract

The art of watermarking and steganography are techniques to hide a secret message in an image or other media. They can be used to protect the privacy and security of medical images and patient data in the healthcare industry. We propose two algorithms for watermarking and steganography based on discrete wavelet transform (DWT), discrete cosine transform (DCT), parallel bacterial foraging optimization (PBFO), and genetic algorithm (GA). These algorithms can embed a secret message in a grayscale medical image of size 512x512 pixels, with high imperceptibility and capacity. Imperceptibility means that the stego image does not have noticeable distortions from the cover image. Capacity means that the stego image can carry as much information as possible without compromising imperceptibility. Algorithm I uses DWT to decompose the cover image into four sub-bands, and DCT to transform each sub-band into a matrix of coefficients. It uses PBFO to select the best coefficients for embedding the message, and GA to embed the message using a binary encoding scheme. Algorithm II is similar to Algorithm I, but it does not use PBFO. It uses GA to select and embed the coefficients directly. We compare the performance of Algorithm I and Algorithm II using peak signal-to-noise ratio (PSNR) and payload. PSNR measures the similarity between two images, where higher values indicate better quality. Payload measures how many bits of information can be embedded per pixel in an image, where higher values indicate more capacity. We use grayscale medical images from a public database as cover images, and random binary strings as secret messages. We vary the length of the secret message from 0.1 to 0.5 bits per pixel. The results show that Algorithm I has better performance than Algorithm II in terms of PSNR and payload. Algorithm I produces less distortion and can embed more information than Algorithm II. Therefore, Algorithm I is more suitable for watermarking and steganography in medical images than Algorithm II. We also discuss some possible applications and challenges of watermarking and steganography in the healthcare domain.