| 摘要 |
Characters read as images at different angles and as mirrored images in accordance with scanning directions are detected by an character recognition procedure, and are converted into erecting images. Such character images have low character recognition rates, therefore, a character image is extracted from the scanned image, a rotated angle and an existence of mirroring in which the recognition rate becomes maximum are detected. Then, the same rotation and mirroring are applied to the scanned image, thereby capable of converting the scanned image into an erecting image.
|
