Unlock instant, AI-driven research and patent intelligence for your innovation.

Text erasing and keying method based on u-shaped residual network

A text erasure and residual technology, applied in the field of image processing, to achieve the effect of avoiding gradient disappearance

Active Publication Date: 2022-04-22
SOUTH CHINA UNIV OF TECH
View PDF7 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] The main purpose of the present invention is to overcome the shortcomings and deficiencies of the prior art, and propose a text erasing and word-cutting method based on a U-shaped residual network. The basic problems of image positioning and image domain transformation can well deal with the text erasure and deduction functions of complex text

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Text erasing and keying method based on u-shaped residual network
  • Text erasing and keying method based on u-shaped residual network
  • Text erasing and keying method based on u-shaped residual network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment

[0066] Such as figure 1 Shown, the present invention, based on the text erasing of U-shaped residual network and the method for digging words, comprises the following steps:

[0067] S1. Construct a training set and a standard answer map, use the existing natural image library in computer vision tasks, combine image RGB channel overlay and mask overlay and other technologies, synthesize natural scene images containing text, and generate subsequent U-shaped residual networks The standard answer map for training, specifically:

[0068] Collect image datasets with rich textures to increase the types of background and text textures, making the trained model more robust; use python language to generate text masks and use computer graphics corrosion deformation to make masks of different shapes , combined with the rendering function of computer graphics to create a variety of rich text textures; use the RGB channel of the image to superimpose natural textures to create text and bac...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

PUM

No PUM Login to View More

Abstract

The invention discloses a text erasing and word-cutting method based on a U-shaped residual network. The method includes the following steps: constructing a training set and a standard answer map, manufacturing a natural scene image containing text, and generating a standard answer map; preprocessing the training set; Extract features, input natural scene pictures containing text into the U-shaped residual network for training, and extract information; image reconstruction, after splicing the output features of multiple U-shaped residual networks, retain the low-frequency information of the image through residual connections , combining the high-frequency information parsed by the autoencoder to output images; deep supervision, optimizing the U-shaped residual network; cyclically training the network until the required standard is obtained. Based on the architecture of U-shaped residual network and automatic encoder, the invention solves the basic problems of text image positioning and image domain transformation, and can deal with text erasure and deduction of complex text.

Description

technical field [0001] The invention belongs to the technical field of image processing, and in particular relates to a text erasing and character-cutting method based on a U-shaped residual network. Background technique [0002] Currently, scene text editing faces two main challenges: text style transfer and background texture preservation. In particular, text styles are composed of multiple factors, such as language, font, color, direction, stroke size, and space angle, making it difficult to accurately capture the complete text style in the source code; at the same time, it is also difficult to maintain the consistency of the editing background, especially When the text appears in some complex scenes, such as menus and signs of street shops. [0003] Autoencoders in deep learning can achieve style transfer in the image domain, and this technique has greatly deepened the research on automatic image editing. However, the low-frequency information of the image will be grea...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to View More

Application Information

Patent Timeline
no application Login to View More
Patent Type & Authority Patents(China)
IPC IPC(8): G06V30/18G06V20/62G06V10/44G06V10/82G06N3/04G06N3/08
CPCG06N3/084G06V20/62G06V30/153G06V10/44G06N3/045
Inventor 许勇余博西黄艳
Owner SOUTH CHINA UNIV OF TECH
Features
  • R&D
  • Intellectual Property
  • Life Sciences
  • Materials
  • Tech Scout
Why Patsnap Eureka
  • Unparalleled Data Quality
  • Higher Quality Content
  • 60% Fewer Hallucinations
Social media
Patsnap Eureka Blog
Learn More

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2025 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com