Line Segmentation Method for Asian Text Using Width Lists
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current line segmentation methods in OCR systems are inefficient for Asian text and combined text, as they often require long computation times due to the lack of clear breaks or pitches between characters, and existing methods are not effective in accurately segmenting characters composed of multiple connected components.
Innovation Solution
A method that determines a starting point, generates a list of potential character widths, applies a classification method to evaluate the likelihood of error, and selects candidates based on threshold values to reduce computation time and improve accuracy, combining line segmentation and character classification into a single process.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional line segmentation methods are used for Asian text, then character segmentation can be achieved, but computation time becomes extremely long
Solution Approach 1:
The patent segments the character string image into multiple candidate regions based on potential character widths, then applies classification only to these candidates rather than evaluating all possible segmentations. This reduces computation time while maintaining segmentation accuracy for Asian text with multiple connected components.
Solution Approach 2:
The patent generates a list of potential character widths and processes only the most likely candidates through full classification, rather than exhaustively analyzing all possible segmentation solutions. This partial processing approach significantly reduces computation time while still achieving accurate segmentation.
2Measurement precision
If multiple segmentation candidates are evaluated to improve accuracy, then segmentation quality increases, but computation time increases
Solution Approach 1:
The patent changes the parameter evaluation approach by using likelihood of error as a filtering criterion. Candidates are sorted by likelihood of error, and only those below a threshold are fully processed. This parameter-based filtering maintains segmentation accuracy while improving processing speed by avoiding exhaustive evaluation.
Solution Approach 2:
The patent skips full classification processing for candidates with high likelihood of error, rushing through the evaluation by discarding them early. This allows the system to focus computational resources on promising candidates, maintaining accuracy for valid segments while speeding up overall processing.
3Ease of operation
If dissection methods are used for Asian characters, then character isolation is attempted, but accuracy decreases because Asian characters consist of multiple connected components
Solution Approach 1:
The patent creates a universal segmentation approach that works for both single-component Latin characters and multi-component Asian characters. By using potential character width lists and likelihood-based filtering rather than dissection methods, the system achieves accurate segmentation across different character types without requiring type-specific processing.
Data Source
AI summary
A line segmentation method which starts with determining a first starting point coordinate and generating a list of potential character widths dependent on a maximum character width stored in a database and on characteristics of the portion of the line of text corresponding to the maximum character width. The method determines a second portion of the line of text corresponding to the first starting point coordinate and the first width on the list of potential character widths. On the second portion, a classification method is applied providing a likelihood of error for the first width and a candidate character. The likelihood of error is compared with a first threshold determined by a trade-off between speed and accuracy, and if the likelihood of error corresponding to the first width is lower than the threshold value, the candidate character is selected as the character meaning that a segment is known.