| 摘要 |
PURPOSE: A method and a device for detecting character frames using a voting process are provided to detect character frames with a high detection accuracy. CONSTITUTION: A gray pixel image of the image scanner(800) is supplied to an edge extracting unit(801). The edge pixel extracting unit(801) extracts edge pixels at edges of lines of the gray pixel image by calculating a differential value of each pixel and determining whether or not the differential value is larger than a threshold value. The extracted edge pixel is supplied to a voting processing unit(802). The voting processing unit(802) performs a voting process upon a voting region which is set in accordance with a gray gradient of the extracted edge pixel. The voting region is expected to include a center of an expected character frame. Note that a voting process increments a voting count by + 1 at each coordinate within the voting region. Generally, since there are a lot of edge pixels, there are a lot of voting regions. Therefore, the voting processing unit(802) carries out a lot of voting processes. A peak detecting unit(803) detects peak points having peak voting counts from all the coordinates by determining whether or not the voting count of each coordinate is larger than a threshold value. The peak points are registered as character frame center candidates. A character frame combination generating unit(804) generates combinations of a predetermined number of the character frame center candidates detected by the peak detecting unit(803). A score calculating unit(805) calculates a score for each of the character frame combinations. The score is dependent upon the voting counts, equi-interval degree and horizontality degree of coordinates of the character frame center candidates of each of the character frame combinations. A character frame determining unit(806) selects one of the character frame combinations having the largest score, and determines whether or not the score of a selected character frame combination is larger that a threshold value. Only when this score is larger than the threshold value, are the character frame center candidates of such a selected character frame combination determined to be character frame centers.
|
