Design And Implementation Of Document Image Watermarking Algorithm Based On Segmenting Image

With the accelerating pace of information technology in modern society, the development of mobile networks will push digitized human application to a new climax. At the same time, security issues become more and more serious. Protecting copyright and tracing the origin of security information are still unsolved problems of human society. DRM which is a mature technology plays a good role in the protection of electronic document security issues. While the protection of paper documents is still in a initial stage of exploration with respect to the protection of electronic documents. Although many scholars have done a lot of research on text watermarking algorithms, but most of them aim at single plain text documents. As a result, there are rare algorithms applied for documents with complex layout. Therefore, it will be of great practical significance to research watermarking algorithms to deal with complex image documents which have complex layout.In this thesis, we first review the basic principles and framework of digital watermarking. We make a detailed analysis of the distortion and attack of printing and scanning processes on images. Meanwhile as plain text images and nature images have different characteristics, we respectively introduce and analysis some algorithms which are resist to printing and scanning operations. Then we present the basic principles of image segmentation and describe the region-based, edge-based and specific theory-based image segmentation algorithms in detail.Second, we study the basic concept of connected component labeling and some commonly used labeling methods. On the basic of original algorithms which need to scan the images for more than one time, we propose an algorithm that only needs one time to label and merge all the connected components. The algorithm also analyzes some special situations of connected components. We use this method to design an algorithm to segment, de-noise and split single character in a document image with complex layout. To be specific, we first use this method labeling all the connected components. Then we make document image segmentation and de-nosing based on the characteristics of connected components and we combine the composition principle of Chinese characters to split them. The experimental results show that this algorithmachieves a good effect on document images.Third, we introduce the invariant-based document watermarking algorithms related to printing and scanning processes and make a deep analysis of the variation range of the “invariants”. Through a better extension of this algorithm, we make it more robust. We also implement the data classification based DCT watermarking algorithm and deeply analyze it. We propose an image-segmentation based watermarking algorithm and carry out it. The algorithm mainly makes use of a quick connected components method to extract text parts and image parts from one document image. Then it utilizes different kinds of watermarking algorithms to add watermark or extract them from documents. From the experiments, we conclude that our algorithm can be successfully applied on document images with complex layout.