Precise grayscale character segmentation apparatus and method
Precise grayscale character segmentation apparatus and method. The precise grayscale character segmentation apparatus comprises an adjustment and segmentation unit for adjusting and segmenting an inputted low resolution text line image undergone coarse segmentation, so as to generate an adjusted character image; a character image binarization unit for generating a binary character image from the character image inputted therein; a noise removal unit for removing noise information in the binary character image generated by the binarization unit; and a final character image segmentation unit for generating a precisely segmented character image from the binary character image from which noise has been removed.
1. A precise grayscale character segmentation apparatus, comprising:
an adjustment and segmentation unit, for adjusting and segmenting single-character images in a low resolution grayscale text line image undergone coarse segmentation which is inputted therein, so as to generate adjusted and segmented character images;
a character image binarization unit, for generating a binary character image from the adjusted and segmented character image inputted;
a noise removal unit, for removing noise information in the binary character image generated by the character image binarization unit; and
a final character image segmentation unit, for generating a precisely segmented character image from the binary character image from which noise information has been removed.
2. The precise grayscale character segmentation apparatus according to claim 1 , further including:
an amplification unit, interposed between the adjustment and segmentation unit and the character image binarization unit, for amplifying the adjusted and segmented character image generated by the adjustment and segmentation unit before the adjusted and segmented character image is inputted into the character image binarization unit.
3. The precise grayscale character segmentation apparatus according to claim 1 , further including:
a character image enhancement unit, interposed between the adjustment and segmentation unit and the character image binarization unit, for enhancing the adjusted and segmented character image generated by the adjustment and segmentation unit before the adjusted and segmented character image is inputted into the character image binarization unit, to make clearer the strokes of the character within the character image.
4. The precise grayscale character segmentation apparatus according to claim 2 , further including:
a character image enhancement unit, interposed between the amplification unit and the character image binarization unit, for enhancing the adjusted and segmented character image amplified by the amplification unit before the adjusted and segmented character image amplified is inputted into the character image binarization unit, to make clearer the strokes of the character within the character image.
5. The precise grayscale character segmentation apparatus according to any of claims 1 to 4 , wherein the adjustment and segmentation unit includes:
a text line direction detection unit, for detecting the direction of the text line which is in the text line image;
a character image size calculation unit, for calculating the size of the character image; and
a character image adjustment unit, for adjusting the character image in response to the detecting result of the text line direction detection unit and the calculating result of the character image size calculation unit, so that all strokes of the character are contained in the adjusted and segmented character image.
6. The precise grayscale character segmentation apparatus according to claim 3 or 4 , wherein the character image enhancement unit includes:
a background pixel value estimation unit, for estimating the background pixel value of an inputted character image; a background removal unit, for removing the background based on the
background pixel value estimated by the background pixel value estimation unit; and a pixel
value enhancement unit, for enhancing pixel value of the character image from which the background has been removed.
7. The precise grayscale character segmentation apparatus according to claim 6 , wherein the background pixel value estimation unit estimates the background pixel value using a histogram based method.
8. The precise grayscale character segmentation apparatus according to claim 6 , wherein the pixel value enhancement unit enhances the pixel value using an S shape function.
9. The precise grayscale character segmentation apparatus according to any of claims 1 to 4 , wherein the noise removal unit includes:
a connected component analysis unit, for analyzing connected components of the binary character image to find out all pixel points of each connected component and calculate the total number of the connected components within the binary character image;
a noise connected component determination unit, for determining whether a connected component is a noise connected component; and
a noise connected component removal unit, for removing the connected component within the binary character image which is determined as a noise connected component by the noise connected component determination unit.
10. The precise grayscale character segmentation apparatus according to claim 9 , wherein
the noise connected component determination unit determines whether a connected component is a noise connected component by the following two conditions:
size of the connected component<size of the binary character image/scale number, and
distance between the boundary of the connected component and the boundary of the binary character image<threshold; and
a connected component is determined as a noise connected component if it satisfies both of the conditions.
11. The precise grayscale character segmentation apparatus according to claim 10 , wherein the scale number is 3 or 4 times the size of the binary character image.
12. A method for precisely segmenting single grayscale characters in a text line image undergone coarse segmentation, comprising:
a computer performing:
adjustment and segmentation adjusting and segmenting single-character images in an inputted low resolution grayscale text line image undergone coarse segmentation, so as to generate adjusted and segmented character images;
character image binarization binarizing the character image processed by the adjustment and segmentation;
noise removal removing noise information in the binary character image generated at the character image binarization; and
final character image segmentation generating a precisely segmented character image from the binary character image from which noise information has been removed.
13. The method according to claim 12 , further including:
amplification amplifying the adjusted and segmented character image generated at the adjustment and segmentation before the adjusted and segmented character image is processed by the character image binarization.
14. The method according to claim 13 , further including:
character image enhancement enhancing the adjusted and segmented character image amplified at the amplification, before the character image binarization, to make clearer the strokes of the character within the character image.
15. The method according to claim 12 , wherein the adjustment and segmentation includes:
text line direction detection detecting the direction of the text line
character image size calculation calculating the size of the character image; and
character image adjustment adjusting the character image in response to the detecting result of the text line direction detection and the calculating result of the character image size calculation unit, so that all strokes of the character are contained within the character image.
16. The method according to claim 14 , wherein the character image enhancement includes:
background pixel value estimation estimating the background pixel value of a character image;
background removal removing the background based on the background pixel value estimated by the background pixel value estimation; and
pixel value enhancement enhancing pixel value of the character image from which the background has been removed.
17. The method according to claim 16 , wherein the background pixel value is estimated by using a histogram based method at the background pixel value estimation.
18. The method according to claim 16 , wherein the pixel value of the character image from which the background has been removed is enhanced using an S shape function at the pixel value enhancement.
19. The method according to any of claims 12 to 18 , wherein the noise removal includes:
connected component analysis analyzing connected components of the binary character image to find out all pixel points of each connected component and calculate the total number of the connected components within the binary character image;
noise connected component determination determining whether a connected component is a noise connected component; and
noise connected component removal removing the connected component within the binary character image which is determined as a noise connected component at the noise connected component determination.
20. The method according to claim 19 , wherein
a connected component is determined whether to be a noise connected component or not at the noise connected component determination by the following two conditions:
size of the connected component<size of the binary character image/scale number, and
distance between the boundary of the connected component and the boundary of the binary character image<threshold; and
a connected component is determined as a noise connected component if it satisfies both of the conditions.