Dot Text Transformation for Robust OCR
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing optical character recognition (OCR) methods for dot text in images face challenges with non-uniform dot spacing, deformed, skewed, and rotated characters, as well as varying contrast and backgrounds, requiring extensive pre-processing and being sensitive to dot pitches and character angles.
Innovation Solution
A method that extracts sets of dots from images based on selection criteria, transforms them into stroked characters by determining first and second candidate distances along specific orientations, and performs model-based character matching without prior segmentation, enabling robust recognition across varying conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional OCR methods are used for dot text, then recognition can be performed, but recognition accuracy deteriorates due to non-uniform dot spacing, deformed, skewed and rotated characters
Solution Approach 1:
The patent transforms the dot text image by changing spatial parameters - specifically by resampling the image at multiple orientations and interpolating dot positions to create uniformly spaced dot representations. This parameter transformation converts variable dot spacing and skewed characters into standardized forms, improving recognition accuracy while maintaining adaptability to various dot text conditions
Solution Approach 2:
The patent performs preliminary actions by pre-processing the dot text image through multiple orientation transformations and dot position interpolations before the actual character recognition occurs. This preliminary standardization of dot spacing and character orientation prepares the data for more accurate recognition without requiring the OCR system to be highly adaptable to each variation
2Reliability
If extensive pre-processing is applied to handle varying contrast and backgrounds, then recognition robustness improves, but device complexity increases
Solution Approach 1:
The patent segments the pre-processing into distinct, modular operations: orientation detection, multi-orientation transformation, dot position interpolation, and uniform spacing resampling. Each segment handles a specific aspect of the variation (contrast, orientation, spacing) independently, improving robustness while keeping the overall system complexity manageable through functional decomposition
3Measurement precision
If dot text is recognized directly without transformation, then processing speed is maintained, but recognition accuracy deteriorates due to non-uniform dot spacing
Solution Approach 1:
The patent creates transformed copies of the original dot text image at multiple orientations with interpolated dot positions. Instead of modifying the original image or requiring complex adaptive recognition, it generates standardized copies that are easier to recognize accurately, maintaining processing efficiency while improving recognition precision through the copied and transformed representations
Data Source
AI summary
A method and apparatus for determining orientation and dot pitch of characters in an image. A statistical neighborhood of a set of dots of an image is determined. The statistical neighborhood includes a set of points and each point is associated with a position and a statistical measure indicative of a likelihood that one or more dots that satisfy a shape and a size criteria are located at that position. A Fast Fourier Transform (FFT) is computed across the set of points of the statistical neighborhood; and based on the FFT of the set of points, a first orientation and a first distance between adjacent dots of characters along the first orientation, and a second orientation and a second distance between adjacent dots of the characters along the second orientation are determined.