Digital Image Watermarking Techniques Using Machine Learning

Abstract

Recently, digital image watermarking develops very fast and applies to many applications, such as military, communication, medical/healthcare, privacy protection, identification, media file archiving, broadcast monitoring, Remote Education and Insurance Companies, Secured E-Voting Systems, Digital cinema, fingerprinting and many federal or state-issued photo IDs such as a driver's license or passport. However, copyright protection, content authentication and ownership identification are the challenging issues of the open channel information. Digital watermarking is technique for inserting digital information into a multimedia document, which can be later extracted or detected for variety of purposes including multimedia security. In addition to providing security, this technique can also be used to recognize the source, owner, distributor or creator of the data. The factors such as Robustness, transparency, capacity, security and computational cost are needed to be consider in the general watermarking system. However, these important factors like robustness, transparency and capacity are hindering to each other. Therefore, we need to balance these factors. In recent years, different machine learning methods such as fuzzy logic, support vector machine (SVM) play an important role to keep the optimal balance between imperceptibility/transparency and robustness. In addition, to enhance the robustness and security of watermark, researchers are using machine learning methods. The thesis work is divided into five different chapters. Chapter wise description of the thesis report is described as follows: