Method and device for locating text area in image

Abstract

The present invention discloses a character area location method and device in an image. The method comprises: constructing a parameter regression model; generating a text space distribution parameter corresponding to an image through the parameter regression model; and reconstructing a character-non-character binary image corresponding to the image according to the text space distribution parameter. The embodiment of the invention further discloses a character area location device in an image. Compared to the prior art, according to the technical scheme of the embodiment, a traditional mode of character area location through the image contour or area features is completely abandoned, and the character area location method and device in an image perform analysis and location of the character area through the deeper hierarchy meaning characteristic and the image text space distribution parameter to avoid the interference of the size, the character font, the color and the language of the image on location so as to have more accurate location and better robustness; and moreover, the method provided by the invention can be applied to various types of images based on the most basic meaning characteristic of the image and has versatility.

Description

technical field [0001] The present invention relates to the technical field of graphic and text processing, and more specifically, to a method and device for locating character areas in an image. Background technique [0002] In online commodity transactions, since consumers cannot see commodities intuitively, commodity images become an important means for merchants to describe commodities to consumers. However, in order to increase the attention of products, some merchants embed false promotional text in product images, which not only causes malicious competition, but also leads to poor consumer experience. Therefore, each e-commerce website checks the text content of the commodity image to form the monitoring of the commodity image. [0003] Usually, the existing process of reviewing text in product images is generally as follows: firstly, locate the text area in the product image to determine the position of the text part in the image; then, refine the text content accor...

AI Technical Summary

Problems solved by technology

[0004] However, the size, font, color, and language of the product image are uncertain, and the text in the product image is easily confused with the complex background of the product image, which brings great interference to the text area in the positioning image, resulting in It is impossible to locate the text area in the image, so that the clear text content cannot be obtained, and the review of product images cannot be automated

Images