| One of the main feature of modern society is digital information. One of the most important media to promulgate information is image. A great deal of information, like weather information, resources distribution of the Earth, medical diagnosis and so on, can be converted into digital images. As the development of the social economy and the technology and science, more and more attention has been focused on information visualization. And as an advanced technology, digital image processing and data compression also plays more and more important roles in society development. Data compression is not only the linchpin of the modern information expressway, the HDTV, the visible TV and the image-text facsimile, but also plays an important role in aviation reconnaissance and remote sensing, resource reconnoitering, biomedicine engineering and other areas. Furthermore, it is also very important in data storage and data transmission.Document image compression is a very important part of data compression. It's widely used in military affair, government, commerce, finance and other areas. Therefore, studies on the theory and application of document compression have significant meanings both in theory and application.Based on plenty of papers, technology reports and dissertations, this dissertation makes some research works on the theories of document image compression, including the image quality assessment model and the content-based compression model. Furthermore, combined with practicality, this dissertation also studies on how to improve the compression performances while decreasing the computation time. The main points are as follows: 1. The compression model of document image is first studied in chapter 2. The image quality of most document images is assessed by human eyes. Therefore, the human visual characters should be considered when selecting the compression scheme. As for the human eyes, good quality of text or graphic region means a clear image, i.e., the spatial resolution is more important than the color resolution; on the other hand, good quality of the continuous-tone picture region, especially the low color resolution (such as 256 colors) images, means colorful images, i.e., the color resolution is more important. Thus, according to JBIG2, document image is first classified into different regions, and then different coder is used according to the characteristics of each region. Because the coder is depended on the features of the region, the compression model in the dissertation is aCDIC (Content-based Document Image Coding) method. 2. The layout analysis of the document image is studied in chapter 3. The LTIB (Low-passed Threshold Image-based Binarization) method is proposed in binarization, and the HTSEC (Hough Transform-based Structure Element Constructing) method is proposed in line extracting. Layout analysis plays an important role in document image compression. Considering that it's quite time-consuming when the layout of gray-level image is analyzed, halftoning is first used to transform the gray-level image into binary mode. And then, a MMS (Mathematical Morphology-based Segmentation) method is used in layout segmentation. A big rectangle structural element is used in image extraction by morphology opening operation. After removing the image regions, the LTIB method is used to binarize the residual gray-level image. In graphic region extraction, the multi-stage Hough transformation is first used in skew detection, and then the HTSEC method is used in extracting lines with different direction. Finally, the text region can be obtained easily.3. In chapter 4, the Gray-level Reducing-based Text Image Coding (GRTIC) method is proposed to code the low-resolution text image, and a new Vector Description (VD) method is proposed to code the graphic region. The text region and the graphic region images after layout segmentation are usually binary images. However, these two regions have their own peculiarity, thus, it's necessary to study the respective coding method for each different r...
|
PDF Full Text Request