期刊名称：International Journal of Computers and Communications
印刷版ISSN：2074-1294
出版年度：2012
卷号：6
期号：4
页码：183-190
出版社：University Press
摘要：Document images are composed of graphics, pictures, text, and background with varying number of colors. Based on the detected number of colors contained in a document image, a new approach for document image segmentation and classification using an Artificial Neural Network (ANN) technique is proposed. The ANN is designed to learn how to recognize some interesting color patterns from the labeled document images. Then, the unlabeled document images are classified according to these color patterns. This approach aims at segmenting the original image content into consistent and homogeneous four regions: picture, graphics, text, and background. In order to achieve better compression ratios, every component is compressed separately using the most appropriate compression technique.
