Document Skew Detection and Correction in Image Scanners
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image reading apparatuses face usability issues due to decreased character recognition accuracy when documents are skewed during feed scanning, requiring users to manually re-read documents or re-process entire bundles, leading to inefficiencies.
Innovation Solution
An image reading apparatus with a conveyance unit, reading unit, and notification unit that detects document tilt and automatically corrects skew based on calculated tilt angles, notifying users of failed readings and executing skew correction to ensure accurate digitalization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If feed scanning is performed without skew detection, then the reading speed is maintained, but the character recognition accuracy decreases due to document tilt
Solution Approach 1:
The system performs skew detection and correction before OCR processing by calculating the tilt angle from the document image and rotating the image to correct the skew. This preliminary action ensures that the document is properly aligned before character recognition, thereby maintaining high recognition accuracy without sacrificing reading speed.
Solution Approach 2:
Instead of using mechanical skew correction mechanisms that would slow down the reading process, the system substitutes mechanical adjustment with digital image processing. The tilt angle is calculated from the image data and corrected through software-based image rotation, maintaining high reading speed while improving character recognition accuracy.
2Measurement precision
If manual verification of OCR results is performed, then character recognition accuracy is improved, but the time required for document processing increases
Solution Approach 1:
The system performs skew correction before OCR processing to ensure that characters are properly aligned, which improves recognition accuracy and reduces the need for manual verification. By addressing the root cause of recognition errors through preliminary image correction, the system minimizes the time users would otherwise spend manually checking and re-reading documents.
3Measurement precision
If skew correction is performed for all documents, then character recognition accuracy is improved, but the processing time increases
Solution Approach 1:
The system uses efficient digital image processing algorithms to perform skew correction rapidly. By calculating the tilt angle from image features and applying software-based rotation, the system achieves accurate skew correction without significant processing time penalty, unlike mechanical correction systems that would add substantial time overhead.
Data Source
AI summary
An image reading apparatus includes a conveyance unit and a reading unit. Based on read image data from the reading unit, a tilt amount of an edge of a document on a leading edge side in a conveyance direction in which the document is conveyed is decided. If the decided tilt amount exceeds a first threshold where a setting that performs document digitalization with respect to the image data, notification is given of information indicating that reading of the document for the document digitalization has failed. If the decided tilt amount is smaller than the first threshold, skew correction is executed for correcting the tilt amount based on the decided tilt amount with respect to the image data, and the document digitalization is executed with respect to the image data for which the skew correction has been executed.