Text image processing method and device

Abstract

The invention provides a text image processing method which comprises: preprocessing a text image to obtain a binary image and a plurality of connected domains included in the binary image; obtaining convex hulls corresponding to the plurality of connected domains and a character zone circumscribing the convex hulls by using a convex hull algorithm; and subjecting the obtained character zone to character segmentation in the horizontal direction to obtain a plurality of character blocks distributed in the binary image; and merging the character blocks according to the heights of the character blocks in the binary image to obtain word blocks included in the text image. In addition, there is provided a text image processing device matching the method. The text image processing method and device can improve the universality and the accuracy of character segmentation.

Description

technical field [0001] The invention relates to the technical field of character recognition, in particular to a text image processing method and device. Background technique [0002] In text image processing, character segmentation plays an extremely important role in the field of character recognition. It is mainly to segment characters at the location of characters on the basis of obtaining the text area of ​​the image. [0003] Existing character segmentation includes projection segmentation, clustering and template matching. Among them, the projection segmentation method will use the binarized image after image preprocessing to determine the area where the character is located through projection; the clustering method uses the connected area of ​​the character, and the character distribution characteristics of the entire page are used to determine the connected area. The character blocks of the region are merged; while the template matching method is mainly applied to ...

AI Technical Summary

Problems solved by technology

[0004] Existing character segmentation methods can segment characters to a certain extent, but they are often limited in practical applications. On the one hand, the projection segmentation method will cause multiple characters to be segmented into one piece when the characters themselves are inclined. , while template matching is less useful, because it can only be used in specific text occasions

[0005] On the other hand, for the clustering method that uses connected regions to achieve character segmentation, if there are strokes in the acquired characters that are broken or glued, it will not be possible to complete the character segmentation

[0006] Therefore, there are many problems in the existing character segmentation, especially in practical applications, and the versatility and accuracy are not high.

Images