OCR Text Position Correction via Tilt Field Function
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current OCR technologies face challenges in accurately correcting text positions due to image distortions caused by shooting errors or paper deformations, leading to misplacement and inconsistent content recognition, especially in scenarios without borders, resulting in a low success rate for automatic correction.
Innovation Solution
A method that determines the slope of each text block in an OCR recognition result, fits a tilt field function using a cubic surface polynomial, and calculates an offset value through discrete integration to correct the text block positions, utilizing a ridge regression algorithm to optimize the tilt field function and minimize differences between slopes.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional OCR text position correction methods are used, then the correction process is simple, but the correction accuracy is low due to image distortions from shooting errors or paper deformations
Solution Approach 1:
The patent transforms the text position correction problem from direct coordinate adjustment to a slope-based mathematical model. By calculating slopes of text blocks and fitting a tilt field function, the system changes the correction parameters from raw coordinates to slope values, enabling more accurate correction of distorted images while maintaining computational feasibility through regression algorithms.
Solution Approach 2:
The patent introduces a tilt field function as an intermediary between the observed text block slopes and the final position correction. This cubic surface function acts as a mediator that models the underlying distortion pattern, allowing the system to infer true text positions by integrating the slope field, thereby improving correction accuracy without requiring complex direct coordinate transformation.
2Reliability
If no border information is available in the image, then the image processing is simpler, but the text position correction success rate decreases due to inability to establish reference frames
Solution Approach 1:
The patent enables the OCR system to perform self-correction by utilizing the text blocks themselves as reference elements. Instead of requiring external border information or manual intervention, the algorithm calculates slopes from the text blocks present in the image, fits a tilt field function to these slopes, and automatically determines the correction transformation, achieving reliable correction without additional reference structures.
Solution Approach 2:
The patent performs preliminary slope calculation and tilt field function fitting before the actual position correction. By pre-establishing the distortion model through regression analysis of text block slopes, the system prepares the correction transformation in advance, ensuring high success rate for automatic correction without requiring iterative adjustments or post-processing interventions.
3Measurement precision
If a cubic surface function is used to represent tilt trend, then the tilt representation is more accurate, but the computational complexity increases
Solution Approach 1:
The patent applies a cubic surface function which provides more parameters than a simple plane model, allowing it to capture complex non-linear distortion patterns. While this exceeds the minimum necessary complexity for simple cases, it provides sufficient accuracy for diverse distortion scenarios. The regression algorithm efficiently estimates the parameters by leveraging the statistical properties of the slope data, balancing accuracy with computational feasibility.
Data Source
AI summary
The present disclosure provides a method for correcting an OCR text position, a storage medium and an electronic device. The method includes: determining a first slope of each text block in an OCR recognition result of a to-be-processed image; fitting a tilt field function in accordance with the first slope of each text block; determining an offset value of each text block in accordance with the tilt field function; and correcting a position of each text block in accordance with the offset value.