Neural Text Matting for Preserving Artistic Fonts and Styles
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing character recognition technologies struggle to preserve personalized features such as artistic fonts and text styles in images like posters and advertisements.
Innovation Solution
A neural network-based text matting method and apparatus that includes a feature extraction network, intermediate processing network, and feature fusion network to extract and preserve text features like font and color, using convolutional and residual blocks to enhance accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If character recognition technology is used to extract text information, then text detection and recognition accuracy is improved, but personalized text features such as artistic fonts and text styles are lost
Solution Approach 1:
The patent segments the text extraction process into two distinct networks: a text detection network for locating text regions and a text matting network for extracting text features. This segmentation allows each network to specialize in one task, enabling accurate text detection while preserving personalized text features through the matting process.
Solution Approach 2:
The patent introduces an intermediary text matting network that acts as a bridge between text detection and final text extraction. This intermediary network processes the detected text regions and generates precise text masks, thereby preserving personalized text features that would otherwise be lost in direct recognition pipelines.
2Productivity
If a simple text extraction method is used, then processing speed is improved, but text feature completeness and quality deteriorate
Solution Approach 1:
The patent performs preliminary text detection before matting extraction, preparing the input for the matting network by identifying text regions of interest. This preliminary action enables the matting network to focus computational resources on text areas only, improving both processing efficiency and text feature completeness.
Solution Approach 2:
The patent transitions from 2D image processing to 3D feature space by incorporating multiple convolutional layers and feature maps in the matting network. This dimensional expansion allows the network to capture complex text features while maintaining processing efficiency through hierarchical feature extraction.
Data Source
AI summary
The present disclosure provides a text matting method and apparatus based on a neural network, a device, and a storage medium. The text matting method based on a neural network includes: processing a first image with a feature extraction network to obtain feature maps, processing the feature maps with an intermediate processing network to obtain intermediate feature maps, processing the intermediate feature maps with a feature fusion network to obtain a second image, wherein the second image includes a text feature extracted from the first image.