Character detection apparatus and method

1. A character detection apparatus, comprising:

a memory; and

processing circuitry configured to:

extract a feature value of an image region including one or more character strings, wherein the feature value is a luminance or a length of the image region in a longitudinal direction;

determine, based on the feature value, which one of a plurality of character detection schemes has a higher character detection accuracy for the image region; and

select a text candidate region detected by the determined character detection scheme if a superimposition degree among a plurality of text candidate regions detected by the plurality of character detection schemes is no less than a first threshold;

wherein

if the superimposition degree is less than the first threshold and if an inclusive relationship exists, a maximum text line candidate having a largest region among superimposed text line candidates is selected as a text line, and

if a ratio of a superimposed region occupying a minimum text line candidate is less than a second threshold value, each of the superimposed text line candidates is selected as a text line.

2. The apparatus according to claim 1 , further comprising:

the one or more character strings are detected using a first character detection scheme and a first text line candidate is obtained indicating a candidate of a region including the one or more character strings; and

the one or more character strings are detected using a second character detection scheme and a second text line candidate is obtained indicating a candidate of a region including the one or more character strings,

wherein a priority with respect to each of the first character detection scheme and the second character detection scheme is determined, a text line candidate detected by one of the first character detection scheme or the second character detection scheme with high priority if the superimposition degree relating to the first text line candidate and the second text line candidate is no less than the first threshold is selected as the text line.

3. The apparatus according to claim 2 , wherein the processing circuitry is further configured to:

connect pixels with similar features between adjacent pixels in the image region to obtain a plurality of connected components;

generate the first text line candidate by combining the connected components in accordance with a similarity of the connected components and a positional relationship between the connected components;

select one or more character candidate regions indicating a character candidate from the image region; and

generate the second text line candidate by combining the one or more character candidate regions.

4. The apparatus according to claim 3 , wherein the processing circuitry is further configured to:

generate the first text line candidate by a line detection using the connected components, and

generate the second text line candidate by a line detection using the one or more character candidate regions.

5. A character detection method, comprising:

extracting a feature value of an image region including one or more character strings, wherein the feature value is a luminance or a length of the image region in a longitudinal direction;

determining, based on the feature value, which one of a plurality of character detection schemes has a higher character detection accuracy for the image region; and

selecting a text candidate region detected by the determined character detection scheme if a superimposition degree among a plurality of text candidate regions detected by the plurality of character detection schemes is no less than a first threshold;

wherein

if the superimposition degree is less than the first threshold, and if an inclusive relationship exists, selecting a maximum text line candidate having a largest region among superimposed text line candidates as a text line, and

if a ratio of a superimposed region occupying a minimum text line candidate is less than a second threshold value, selecting each of the superimposed text line candidates as a text line.

6. The method according to claim 5 , further comprising:

first detecting the one or more character strings using a first character detection scheme and obtaining a first text line candidate indicating a candidate of a region including the one or more character strings;

second detecting the one or more character strings using a second character detection scheme and obtaining a second text line candidate indicating a candidate of a region including the one or more character strings;

determining a priority with respect to each of the first character detection scheme and the second character detection scheme;

integrating the first text line candidate and the second text line candidate; and

selecting, as the text line, a text line candidate detected by one of the first character detection scheme or the second character detection scheme with high priority if the superimposition degree relating to the first text line candidate and the second text line candidate is no less than the first threshold.

7. The method according to claim 6 , wherein the first detecting comprises:

connecting pixels with similar features between adjacent pixels in the image region to obtain a plurality of connected components; and

first generating the first text line candidate by combining the connected components in accordance with a similarity of the connected components and a positional relationship between the connected components, and

the second detecting comprises:

selecting one or more character candidate regions indicating a character candidate from the image region; and

second generating the second text line candidate by combining the one or more character candidate regions.

8. The method according to claim 7 , wherein the first generating generates the first text line candidate by a line detection using the connected components, and the second generating generates the second text line candidate by a line detection using the one or more character candidate regions.

9. A non-transitory computer readable medium including computer executable instructions, wherein the instructions, when executed by a processor, cause the processor to perform a method comprising:

extracting a feature value of an image region including one or more character strings, wherein the feature value is a luminance or a length of the image region in a longitudinal direction;

determining, based on the feature value, which one of a plurality of character detection schemes has a higher character detection accuracy for the image region; and

selecting a text candidate region detected by the determined character detection scheme if a superimposition degree among a plurality of text candidate regions detected by the plurality of character detection schemes is no less than a first threshold;

wherein

if the superimposition degree is less than the first threshold, and if an inclusive relationship exists, selecting a maximum text line candidate having a largest region among superimposed text line candidates as a text line, and

if a ratio of a superimposed region occupying a minimum text line candidate is less than a second threshold value, the selecting each of the superimposed text line candidates as a text line.

10. The medium according to claim 9 , further comprising:

first detecting the one or more character strings using a first character detection scheme and obtaining a first text line candidate indicating a candidate of a region including the one or more character strings;

second detecting the one or more character strings using a second character detection scheme and obtaining a second text line candidate indicating a candidate of a region including the one or more character strings;

determining a priority with respect to each of the first character detection scheme and the second character detection scheme;

integrating the first text line candidate and the second text line candidate; and

selecting, a text line candidate detected by one of the first character detection scheme or the second character detection scheme with high priority if the superimposition degree relating to the first text line candidate and the second text line candidate is no less than the first threshold.

11. The medium according to claim 10 , wherein the first detecting comprises:

connecting pixels with similar features between adjacent pixels in the image region to obtain a plurality of connected components; and

first generating the first text line candidate by combining the connected components in accordance with a similarity of the connected components and a positional relationship between the connected components, and

the second detecting comprises:

selecting one or more character candidate regions indicating a character candidate from the image region; and

second generating the second text line candidate by combining the one or more character candidate regions.

12. The medium according to claim 11 , wherein the first generating generates the first text line candidate by a line detection using the connected components, and the second generating generates the second text line candidate by a line detection using the one or more character candidate regions.