Text Line Curvature Correction for OCR Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Text images captured by mobile devices and other imaging equipment often suffer from perspective-induced and curved-page-surface-induced curvature, which degrades the performance of optical-character recognition systems by increasing erroneous character recognition and failure in producing accurate text encoding.
Innovation Solution
The method involves identifying the outline of a text-containing page, determining the centroids and inclination angles of symbol aggregations, constructing a model for perspective-induced curvature, and using an inclination-angle map to assign local displacements to pixels, thereby straightening text lines in the image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If text images are captured by mobile devices for convenience, then ease of operation is improved, but perspective-induced curvature and distortion occur degrading recognition accuracy
Solution Approach 1:
The patent applies preliminary action by detecting symbol aggregations and constructing a curvature model before optical character recognition is performed. The method identifies the outline of text-containing pages, determines centroids and inclination angles of symbol aggregations, and builds a perspective-induced curvature model in advance to correct distortions before the actual recognition process, thereby improving measurement precision while maintaining ease of operation
Solution Approach 2:
The patent applies parameter changes by transforming the curvature model into an inclination-angle map that assigns local displacement parameters to pixels. This changes the geometric parameters of the image by calculating displacement based on inclination angles and applying transformations to straighten curved text lines, thereby correcting perspective-induced distortion while preserving the convenience of mobile device capture
2Productivity
If automated optical-character recognition is applied directly to distorted images, then processing speed is improved, but reliability of character recognition deteriorates due to curvature
Solution Approach 1:
The patent applies segmentation by dividing the text-containing page into symbol aggregations (words and word fragments) and identifying their centroids and inclination angles. This segmentation allows the curvature model to be constructed based on discrete symbol groups rather than processing the entire image at once, enabling efficient correction while maintaining high reliability in character recognition by addressing distortion at the symbol level
Solution Approach 2:
The patent introduces an intermediary curvature model and inclination-angle map between the distorted image and the optical character recognition process. This intermediary structure serves as a mediator that transforms the distorted image into a corrected version by assigning local displacements based on inclination angles, thereby improving recognition reliability without significantly impacting processing speed
3Measurement precision
If curvature correction is applied to straighten text lines, then measurement precision is improved, but device complexity increases due to additional processing steps
Solution Approach 1:
The patent applies self-service by using the image data itself to construct the curvature model. The method determines centroids and inclination angles of symbol aggregations directly from the input image, and uses these extracted features to build the perspective-induced curvature model without requiring external calibration data or complex reference systems. This self-service approach improves text line straightness while limiting the increase in system complexity
Data Source
AI summary
The current document is directed to methods and systems that acquire an image containing text with curved text lines to generate a corresponding corrected image in which the text lines are straightened and have a rectilinear organization. The method may include identifying a page sub-image within the text-containing image, generating a text-line-curvature model for the page sub-image that associates inclination angles with pixels in the page sub-image, generating local displacements, using the text-line-curvature model, for pixels in the page sub-image, and transferring pixels from the page sub-image to a corrected page-sub-image using the local displacements to construct a corrected page sub-image in which the text lines are straightened and in which the text characters and symbols have a rectilinear arrangement.