Layout-Aware OCR Using Text Line and Word Unit Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional optical character recognition (OCR) technologies struggle with accurately aligning recognized characters to the layout of text in images, leading to errors in text interpretation.
Innovation Solution
A character recognition system utilizing a deep learning model to extract character area, inter-character space, interline scale, and orientation information, followed by a text line recognition unit to align characters based on these features, and a layout analysis unit to generate paragraph and line number information, ensuring accurate text alignment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional OCR technology is used to recognize characters in images, then character recognition can be performed, but the recognized characters are not aligned to correspond to the layout of the text, causing errors in text interpretation
Solution Approach 1:
The patent segments the text recognition process into distinct components: character recognition unit, text line recognition unit, and layout analysis unit. Each unit processes specific aspects (characters, lines, layout structure) separately and integrates results to preserve both character accuracy and layout information simultaneously.
Solution Approach 2:
The patent adds layout dimension to traditional OCR by introducing layout analysis that operates in the spatial dimension. Instead of only recognizing characters in isolation, the system analyzes text line positions, orientations, and hierarchical structures, transforming 2D image coordinates into structured layout information that preserves spatial relationships.
2Measurement precision
If deep learning technology is introduced to improve character recognition rate and accuracy, then recognition accuracy improves, but the system becomes more complex
Solution Approach 1:
The patent divides the complex deep learning system into modular units with specialized functions. Each unit (character recognition, text line recognition, layout analysis) uses appropriate deep learning models for its specific task, avoiding the need for one overly complex model to handle all aspects, thus managing system complexity through functional segmentation.
Solution Approach 2:
The patent employs deep learning models that serve multiple functions across different units. The same deep learning framework and techniques are applied in character recognition, text line recognition, and layout analysis, creating a universal approach that reduces overall system complexity through methodological consistency while maintaining high accuracy.
3Measurement precision
If text line recognition is performed based on character area information, interline scale information, and orientation information, then text line recognition accuracy improves, but the processing time increases
Solution Approach 1:
The patent performs preliminary extraction of character area information, interline scale information, and orientation information in the character-related information extraction unit before text line recognition. By pre-processing and organizing these features, the text line recognition unit receives ready-to-use data, reducing its processing time while maintaining high accuracy through the use of pre-computed spatial features.
Solution Approach 2:
The patent replaces traditional mechanical or rule-based text line grouping methods with deep learning-based recognition. The deep learning model automatically learns patterns in character spatial relationships, interline scales, and orientations, substituting complex mechanical processing rules with a more efficient neural network approach that achieves high accuracy with optimized processing time.
Data Source
AI summary
A character recognition system includes: a character-related information extraction unit configured to include a deep learning model trained to extract character area information, inter-character space area information, interline scale information of each character, and orientation information of each character from an image including text; a word unit division recognition unit configured to obtain word division information obtained by dividing characters included in the image into word units based on the character area information and inter-character space area information; a text line recognition unit configured to recognize text lines in the image based on the character area information, interline scale information, and orientation information; a layout analysis unit configured to obtain layout information of the text included in the image based on the recognized text lines; and a character recognition unit configured to recognize each of the character included in the image and obtain text data in which the recognized characters are aligned based on the word division information and the layout information.