Text Line Image Splitting for Mixed Font OCR Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional optical character recognition (OCR) systems struggle with text line images containing text with differing fonts, sizes, and vertical positions, leading to inaccurate normalization and recognition, as they require text lines to be of a fixed size and consistent position.
Innovation Solution
A method that identifies zones with differing fonts in text line images and splits them into segments based on OCR confidence measurements and text height confidence, allowing for separate normalization and recognition of each segment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If conventional OCR systems process text line images with varying fonts and sizes, then the system can handle diverse document formats, but the recognition accuracy deteriorates due to inability to normalize text properly
Solution Approach 1:
The text line image is divided into multiple text segments based on font size variations. Each segment containing text with similar font characteristics is processed separately through normalization and OCR, then results are combined. This segmentation allows the system to adapt to diverse font styles while maintaining high recognition accuracy for each segment.
Solution Approach 2:
Different normalization parameters are applied to different regions of the text line image based on local font characteristics. Each text segment receives customized processing parameters appropriate to its specific font size and style, rather than applying uniform parameters to the entire line, thereby improving recognition accuracy for mixed-font documents.
2Productivity
If text line images are normalized to fixed size before OCR, then OCR processing becomes simpler and faster, but text lines with varying font sizes cannot be properly normalized
Solution Approach 1:
Instead of attempting to normalize the entire text line to a fixed size, the method segments the text line into multiple parts based on font size thresholds. Each segment is then normalized independently to a standard size, enabling both efficient OCR processing and accurate representation of the original text characteristics.
Solution Approach 2:
The normalization process becomes dynamic by adapting the target size and parameters based on the detected font characteristics of each segment. Rather than a static fixed-size normalization, the system dynamically adjusts normalization parameters to preserve the relative size relationships among different font elements while enabling standardized processing.
3Device complexity
If the entire text line image is processed as a single unit, then the processing workflow is simpler, but recognition accuracy deteriorates when different fonts are present
Solution Approach 1:
The text line image is automatically segmented into multiple sub-images based on detected font size variations. Each segment is then processed through the normalization and OCR workflow separately, with results aggregated to produce the final recognition output. This maintains a relatively simple workflow structure while significantly improving accuracy for mixed-font text.
4Productivity
If text lines with different fonts are normalized using a single parameter set, then the normalization process is faster, but recognition accuracy decreases for non-uniform text
Solution Approach 1:
The text line is segmented into multiple regions based on font size criteria before normalization. Each segment is then normalized using parameters appropriate to its specific font characteristics, rather than applying a single parameter set to the entire line. This segmented approach maintains efficient processing speed while achieving accurate normalization for each font type.
Solution Approach 2:
The normalization process applies different parameters to different regions of the text line based on local font properties. Each text segment receives customized normalization parameters tailored to its specific font size and style, enabling accurate normalization of non-uniform text while maintaining reasonable processing efficiency through localized operations.
Data Source
AI summary
A method for splitting text line images includes receiving a text line image and identifying that the text line image comprises a plurality of zones, wherein each zone includes text whose font differs from the text of adjacent zones. The method further includes selecting a splitting position between multiple zones and splitting the text line image at the splitting position into a plurality of image segments, wherein each image segment contains at least one zone of the text line image and performing optical character recognition on each image segment to recognize a text segment of the image segment. In certain implementations, the method further includes generating one or more confidence measurements and selecting a splitting position that corresponds to a large gradient in the confidence measurement.