OCR Text Graph Structuring for Overlapping Scenes
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current OCR systems face challenges in recognizing text overlapping areas due to difficulties in distinguishing overlapping text, leading to low recognition rates and information loss, even with noise reduction methods like background removal.
Innovation Solution
The method involves processing text as graph-structured data, capturing endpoints, turning points, and intersections as nodes, and lines as edges, constructing a graph template library, and converting overlapping text regions into topology graphs for subgraph segmentation and recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional background removal methods are used to eliminate overlapping text, then the foreground text recognition is improved, but background text information is lost
Solution Approach 1:
The patent segments the overlapping text region into multiple independent text layers by detecting text contours and hierarchical relationships. Each layer is processed separately through graph template matching, allowing both foreground and background text to be recognized independently without information loss
Solution Approach 2:
The patent transforms the 2D overlapping text image into a hierarchical graph structure with multiple layers. By adding the dimension of text layer hierarchy, the system can distinguish and process overlapping text at different depths, enabling simultaneous recognition of both foreground and background text
2Productivity
If OCR algorithms attempt to recognize overlapping text directly, then text recognition is attempted, but recognition accuracy deteriorates due to text overlap
Solution Approach 1:
The patent divides the overlapping text region into multiple non-overlapping text layers by detecting text contours and establishing hierarchical relationships. Each layer contains distinct text characters that can be recognized independently, eliminating the interference caused by text overlap
Solution Approach 2:
The patent introduces graph template matching as an intermediary processing step between image input and text recognition. The graph template library serves as a mediator that matches text contours at different hierarchical levels, enabling accurate recognition of overlapping text by comparing against predefined graph templates
3Measurement precision
If noise reduction methods are applied to remove background, then foreground text clarity is improved, but overall text information completeness deteriorates
Solution Approach 1:
The patent segments the image into multiple text layers based on contour detection and hierarchical relationships, allowing each layer to be processed independently. This segmentation enables foreground text to be clearly recognized while preserving background text information in separate layers
Solution Approach 2:
Instead of removing background to improve foreground recognition, the patent inverts the approach by treating both foreground and background as valid text layers that should be preserved and processed. The hierarchical graph structure allows simultaneous processing of multiple layers without eliminating any text information
Data Source
AI summary
Embodiments of the present disclosure provide systems and methods for implementing enhanced Optical Character Recognition (OCR) of text overlapping scenes through text graph structuring. Text graph structuring is performed to provide a graph data structure for each data character or letter of multiple letters and a library of graph templates from graph structured data of each of the multiple letters. Text graph structuring is performed to convert visual content of an identified overlapping text image region to an overlapping text topology graph. The overlapping text topology graph is split into multiple subgraphs using the graph template library to match recognizable letters in the overlapping text.