A new algorithm separating text strings from map images

To enter geographic data into computers, people desire to have a system which can scan a paper-based map and feed the raster map image into a computer. Then the computer should automatically extract the geographic information from the map image by image analysis and understanding techniques. A new algorithm which can separate text strings of various fonts, sizes, and styles from other graphic data (line drawing) in the map image is proposed. Numerals are automatically distinguished from Chinese characters based on curved strokes. Disconnected components or parts in a Chinese character can be grouped into meaningful character strings. Experimental results of several test images show the effectiveness of this algorithm