Method and device for detecting text in native electronic image

Abstract

The invention discloses a method and device for detecting a text in a native electronic image, and the method comprises the steps of determining M binarization thresholds based on a gray histogram of a to-be-detected image; processing the to-be-detected image based on the M binarization thresholds to obtain M binarization images; and processing the M binary images to obtain M groups of text areas, wherein each group of text areas corresponds to one binary image. According to the method, by respectively applying the plurality of binarization thresholds to process the to-be-detected image, the plurality of binary images are obtained, and after the plurality of groups of text areas are obtained through the plurality of binarized images, the target text area of the to-be-detected image is determined based on the multiple text areas, and the mutual compensation exists among the multiple text areas, so that the probability that the background and the text areas are fused due to the fact that the background color is similar to the color of the text areas can be reduced, the probability that the text areas are missed is reduced, and the text area detection accuracy is improved.

Description

technical field [0001] The invention relates to the technical field of image processing, in particular to a method and device for detecting text in an original electronic image. Background technique [0002] Original electronic pictures are processed or synthesized by computer, not directly produced by cameras, such as advertising pictures, pictures on emails or web pages, etc. Usually, the text on these pictures is fair print, but the text is superimposed on a complex background. In order to distinguish the text from the complex background, the current mainstream method is to select an appropriate threshold and use it to process the image to obtain a black and white image, and then pass MSER (Maximally Stable Extremal Regions, maximum extreme stable region) and SWT (Stroke Width Transform, stroke width transformation) algorithm processes black and white pictures to determine the text area. However, in the case where the background color is similar to the color of the text...

AI Technical Summary

Problems solved by technology

[0003] The embodiment of the present invention provides a text detection method and device to solve the problem that the text area detection is not accurate enough when the background color is similar to the text area color

Images