Genetic algorithm for selecting relevant regions in digital watermarking scheme for 2D/3D medical images

Abstract

In the paper, we have proposed an algorithm of selecting relevant regions for watermark embedding in 2D/3D medical images. Different types of watermarks, such as region of interest, patient electronic record, and fragile watermark, require varied quality after extraction from the watermarked image. First, we select the relevant regions in a host image and second, the genetic algorithm is employed to search the best coefficients for embedding using the corresponding transform. The objective function is calculated under the assumption that a watermarked image was damaged by attacks. In this study, we simulated the simplest types of attacks, such as JPEG compression, translation, and rotation. In the case of 3D images, we consider 3D image as a set of sliced images according to the DICOM standard. This means that the region of interest appears in each sliced image. We suggest an automatic segmentation using the method of local and global intensity fitting with the following Bregman iterative procedure.