Document Skew Detection Using Character Center Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing apparatuses face challenges in accurately detecting the skew of a document image, especially when the document is placed in a skewed state or fed by an automatic document feeder, as they rely on edge detection or corner point positioning, which can be unreliable due to lack of edge shadows or similar background colors.
Innovation Solution
The apparatus includes a binary image generation portion, a specific character position detection portion, and a skew angle specifying portion to binarize the image, detect center positions of specific characters, and calculate the skew angle based on their alignment direction, allowing for accurate skew detection without relying on edge detection or background color differentiation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If edge detection method is used to detect document skew, then the detection process is simple, but the detection accuracy deteriorates when edge shadows are absent or background colors match the document
Solution Approach 1:
The patent introduces an intermediary element (edge shadow or color difference) to facilitate skew detection. When direct edge detection fails due to absent shadows or matching background colors, the system uses the intermediary of color difference between document and background to enable accurate skew detection through corner point positioning
Solution Approach 2:
The patent changes the detection parameter from relying solely on edge shadows to utilizing color difference parameters. By detecting the color difference between the document and background, the system can identify document edges and corner points even when traditional edge shadow detection fails, thereby maintaining high skew detection accuracy
2Measurement precision
If corner point positioning based on background color differentiation is used, then skew detection accuracy improves, but the method fails when background colors match the document
Solution Approach 1:
The patent implements a dynamic detection approach that adapts to different document-background color relationships. The system automatically switches between edge shadow detection and color difference detection methods based on the actual imaging conditions, ensuring reliable skew detection regardless of whether the background color matches the document
3Device complexity
If traditional edge detection or corner point positioning is used, then the processing is straightforward, but accuracy deteriorates in skewed document scenarios
Solution Approach 1:
The patent implements a feedback mechanism where the system evaluates the success of initial skew detection attempts and automatically adjusts its approach. If corner point positioning fails to identify valid points (indicating skewed document placement), the system provides feedback to trigger alternative detection methods, ensuring accurate skew detection while maintaining processing efficiency
Data Source
AI summary
A binary image generation portion binarizes a scanned image and generates a binary image. A specific character position detection portion (a) specifies rectangles that circumscribe a plurality of characters, respectively, in the binary image, (b) detects a plurality of specific characters having longitudinal lines or transverse lines at centers of the specified circumscribing rectangles, and (c) detects, as positions of the plurality of specific characters, center positions of the longitudinal lines or the transverse lines, within a predetermined range, to be detected, in a main scanning direction or a sub-scanning direction. A skew angle specifying portion specifies an alignment direction in which the plurality of specific characters are aligned, based on the positions of the plurality of specific characters detected in the range to be detected, and specifies a skew angle of a document image in the scanned image, based on the specified alignment direction.