A New Approach for Processing Compressed Data

随着数字信息量的快速增长，高效存储、传输或处理数字数据的能力成为一个重要问题。在处理数字信息方面的任何进一步改进，超越了存储容量和通信带宽的技术限制，都应该来自于处理压缩数据。在本项目的第一部分，研究了文档图像理解和OCR中常见的操作在CCITT-G4压缩数据上的执行情况。这些操作包括：斑点着色、细化和形态技术。使用和不使用渐进式编码的JBIG编码文档的直接操作是本项目的另一个目标。针对JBIG数据，研究了倾斜估计和连通分量算法。此外，还实现了压缩后的二值文档图像和工程图纸的图像处理算法。该项目的第二部分涉及开发一种新的源代码的可行性，该源代码允许对OCR生成的文本数据进行快速字符串匹配(精确和近似)。此外，还提出了一种新的无损二值图像压缩方案，该方案允许对图像旋转进行快速的倾斜估计。在该项目下开发的算法应有助于有效利用信息、计算和电信技术中的存储和带宽。

As the amount of digital information increases rapidly, the ability to store, transmit, or process digital data efficiently becomes a significant issue. Any further improvement in handling digital information, beyond technological limits in storage capacity and communication bandwidth, should come from manipulating compressed data. In the first part of this project, execution of common operations in document image understanding and OCR on CCITT-G4 compressed data is examined. These operations include: blob coloring, thinning, and morphological techniques. Direct manipulation of JBIG-coded documents with and without progressive coding is another objective of this project. For JBIG data, skew estimation and connected component algorithms are investigated. In addition, image processing algorithms for compressed binary document images and engineering drawings are implemented. The second part of this project deals with the feasibility of developing a new source code that allows quick string matching (exact and approximate) for OCR-generated text data. Furthermore, a new lossless binary image compression scheme that allows fast skew estimation for image rotation is developed. Algorithms developed under this project should contribute to efficient utilization of storage and bandwidth in information, computation and telecommunication technologies.