Skew Correction for Document Images Using Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image-correcting apparatuses are unable to effectively correct gentle skew in documents during conveyance by deforming only necessary parts of the image.
Innovation Solution
An image-processing apparatus that includes units for detecting edge-skew and content-skew occurrence positions, allowing for the correction of accumulated skew in the document image by geometrically transforming the document rear-part image to align with approximate straight lines, thereby correcting skew without altering the front-part image.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If conventional image-correcting apparatuses are used, then the apparatus structure is simple, but the apparatus cannot effectively correct gentle skew in documents during conveyance
Solution Approach 1:
The patent divides the document image into a front-part image and a rear-part image based on the detected skew occurrence position. The front-part image (from the leading end to the skew position) is kept unchanged, while only the rear-part image (from the skew position to the trailing end) is corrected for skew. This segmentation allows selective correction of only the necessary portion, improving reliability without requiring complete redesign of the entire image processing system.
Solution Approach 2:
The patent applies different processing qualities to different parts of the document image. The front-part image maintains its original quality and characteristics, while the rear-part image receives specialized skew correction processing. This local differentiation enables effective skew correction in the affected region without altering the unaffected region, resolving the contradiction between correction capability and overall system complexity.
2Reliability
If the entire document image is corrected for skew, then the skew correction is comprehensive, but the front-part image is unnecessarily altered
Solution Approach 1:
The patent segments the document image into two distinct parts: a front-part image from the leading end to the skew occurrence position, and a rear-part image from the skew position to the trailing end. This segmentation enables the system to apply skew correction only to the rear-part image where skew actually occurs, while preserving the front-part image unchanged. Consequently, skew correction accuracy is maintained in the affected region, and image data integrity is preserved in the unaffected region.
Solution Approach 2:
The patent extracts only the necessary portion (rear-part image) that contains the skew distortion and applies correction to that extracted segment. The front-part image, which does not contain skew, is excluded from the correction process. This extraction approach ensures that skew correction accuracy is achieved where needed while avoiding unnecessary alteration of the front-part image, thus preventing information loss.
3Reliability
If skew correction is applied to the entire image, then all skew is corrected, but processing time increases
Solution Approach 1:
The patent segments the document image into a front-part image and a rear-part image based on the detected skew occurrence position. By identifying where skew begins (at the skew occurrence position), the system can segment the image such that only the rear-part image requiring correction is processed. This segmentation dramatically reduces the processing area and consequently the processing time, while maintaining complete skew correction for the affected region.
Solution Approach 2:
Instead of applying skew correction to the entire image (excessive action), the patent applies correction only to the necessary rear-part image (partial action). This partial correction approach achieves complete skew correction for the affected region while significantly reducing processing time compared to correcting the entire image, resolving the contradiction between correction completeness and processing efficiency.
Data Source
AI summary
According to the present disclosure, edges of a document image contained in a read image read by an image-reading apparatus including a paper path are detected, a candidate edge-skew occurrence position is detected, a candidate content-skew occurrence position is detected, an accumulated-skew occurrence position is detected based on the candidate edge-skew occurrence position and the candidate content-skew occurrence position, skew of a document rear-part image rearward from the accumulated-skew occurrence position in the conveyance direction is corrected, and image data corrected for accumulated skew is acquired.