Chinese character image restoration algorithm considering Chinese character pattern structure characteristics

Abstract

As a branch of image restoration, character restoration is widely paid attention in recent years, but most of the character restoration at present aims at English characters with simple structures, less research is carried out on Chinese characters with complex structures and diversification, and meanwhile, most of the image restoration work including the character restoration at present aims at damaged images. At present, there is no more research on restoration of a smeared image. The invention provides a character repairing algorithm using a font structural feature attention network. The method can be used for effectively and structurally repairing damaged Chinese character images, painted Chinese character images or Chinese character images with the damaged Chinese character images and painted Chinese character images. Wherein the font structural feature attention network considers that the font structural feature attention network is different from English, Chinese characters are various in variety and different in structure, and if a traditional image restoration technology is used, specific restoration is difficult to carry out according to different structural features of the Chinese characters, so that the font structural feature attention network is developed from the structure of the Chinese characters. A font structural feature attention network for performing specific repair on different Chinese characters is provided.

Description

technical field [0001] The technical fields involved in the present invention include computer vision, artificial intelligence, image restoration, etc., and is a Chinese character image restoration algorithm considering the structural features of Chinese characters. Background technique [0002] Image inpainting, as one of the hot research directions in the field of computer vision, aims to restore the pixel features of damaged parts in incomplete images, and plays a key role in many computer vision application fields. In recent years, text inpainting, a sub-branch of image inpainting, has received continuous attention. However, because of the particularity of the text, a slight structural difference between the repair result and the real image will result in a greatly reduced visual effect, which is also the difficulty of text repair. [0003] In recent years, image inpainting technology based on deep learning network has developed rapidly. At the 2016 CVPR conference, Pa...

AI Technical Summary

Problems solved by technology

[0008] In order to repair the damaged or smeared Chinese character images in different real scenes, such as inscriptions, ancient books, calligraphy works and other cultural relics, the present invention solves the problems of complex structure of Chinese characters, large number, and many similar characters that affect the repair effect

The main repair objects of the present invention are damaged or smeared Chinese characters with complex structures. Inspired by the attention mechanism, the present invention takes the glyph structure of Chinese characters as inspiration, and proposes a glyph structural feature attention network TAN to solve the glyph structure of Chinese characters. For complex and diverse problems, image features are segmented and selected to remove features that have nothing to do with the glyph structure and features that hinder glyph repair, so as to obtain better repair results

Images