Image processing device and image processing method

Abstract

When there is a significant difference in the difference between the number of character pixels having a gray scale level having the largest number of character pixels and the number of character pixels having a gray scale level having the second largest number of character pixels, an image processing device selects, as an output image, character candidate elements obtained after removing non-character elements in the gray scale level having the largest number of character pixels. When there is not a significant difference in the difference between the number of character pixels having the gray scale level having the largest number of character pixels and the number of character pixels having the gray scale level having the second largest number of character pixels, the image processing device selects, as an output image, character candidate elements obtained after removing non-character elements in the gray scale level having the smallest number of edge pixels.

Description

technical field [0001] Embodiments of the present invention relate to an image processing device and an image processing method for extracting characters from image data. Background technique [0002] Conventionally, there is known a technique for extracting characters by removing the trimmed parts when image data includes trimmed characters. As a specific method, there is the following technology: After converting the image data into a binary image, the characters are cut out according to each grayscale, and it is determined whether the cut out characters are normal characters or hollow characters. Image correction for conversion to normal characters, and characters are obtained. In addition, there is also a technique of performing character recognition on image data by referring to a plurality of character recognition dictionaries, calculating the reliability of characters as the respective execution results, and selecting a recognition result with higher reliability. ...

AI Technical Summary

Problems solved by technology

[0009] However, when extracting characters by conventional image processing, it may be difficult to cut out characters when the edges are thick or when characters touch each other.

Also, in the case of using character recognition dictionaries, calculations are performed by combining multiple character recognition dictionaries, so the amount of calculation tends to increase.

Images