| Co-Investigator(Kenkyū-buntansha) |
OKADA Yoshihiro Kyoto University, Faculty of Engineering, 工学部, 教務職員 (30127063)
MINOH Michihiko Kyoto University, Faculty of Engineering, 工学部, 助手 (70166099)
ARIKI Yasuo Kyoto University, Faculty of Engineering, 工学部, 助手 (10135519)
|
|---|
| Budget Amount *help |
¥20,700,000 (Direct Cost: ¥20,700,000)
Fiscal Year 1986: ¥6,700,000 (Direct Cost: ¥6,700,000)
Fiscal Year 1985: ¥14,000,000 (Direct Cost: ¥14,000,000)
|
|---|
| Research Abstract |
The purpose of this research is to develop a drawing image processor which can translate words or phrases contained in drawing images from one language to another, according to human specification of image size, line width,character font sixe, etc. Drawing images are, at first, fed from a image scanner to data flow processor. The data flow processor extracts the features such as connected regions, vertical or horizontal lines. The features are,then, transmitted from the data flow processor to a host processor. The host processor identifies the character string and extracts the connecting points. These structural data of input drawing images are transformed according to the specification. Then, an output image is generated using the transformed structural data and the input image. In this drawing image processor, segmentation and labeling of constituents and translation, transformation, editing and fair printing of drawing images are included.
At 1985, we developed the first version of the drawing image processor. At the year 1986, we improved and intensified the processor to a practical level in the following points.
1. Error occurs frequently in structural analysis when image quality decreases due to contact of characters to lines, or discontinuity of lines. To solve this problem, knowledge base system was constructed by injecting the knowledge concerning "line", "character", "dotted line" and "arrow".
2. Graphics software was developed to increase the quality of output drawing images by using CAD techniques. Fair printing is obtained by this software.
3. Subpattern recognition of chinese character in the low quality drawing is studied.
|
