Document Edge Detection Using Dynamic Reference Line Selection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies face challenges in accurately identifying document information, such as edges and inclination, when a part of the document extends beyond the boundaries of the read image, leading to difficulties in edge detection and inclination calculation.
Innovation Solution
A computer-readable medium storing a program that executes image processing operations, including read image data acquisition, line identification, and document information identification using reference lines formed from candidate lines. The program determines whether an inside reference line group or an end reference line group is selectable based on the ratio of end lines to total candidate lines, allowing for appropriate identification of document information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional edge detection methods (Hough transform, least-square technique) are used to calculate document inclination, then the inclination can be detected when the document is fully visible, but the detection accuracy deteriorates when a part of the document extends beyond the read image boundaries
Solution Approach 1:
The patent dynamically switches between two different inclination detection methods based on the visibility condition of the document. When the document is partially visible (extends beyond read image boundaries), it uses a geometric calculation method based on visible edge positions. When the document is fully visible, it uses conventional methods like Hough transform. This dynamic adaptation resolves the contradiction by selecting the appropriate method for each situation.
Solution Approach 2:
The patent changes the detection parameters and methodology based on the document visibility state. It determines whether the document extends beyond boundaries and adjusts the inclination calculation approach accordingly - using coordinate-based geometric calculations for partial visibility versus pixel-based transform methods for full visibility. This parameter adaptation enables accurate detection in both scenarios.
2Area of stationary object
If the document extends beyond the read image boundaries, then more of the document content is captured, but the edge detection and inclination calculation become inaccurate or impossible
Solution Approach 1:
The patent segments the document edge detection process into two distinct cases: partial visibility and full visibility. For partial visibility, it identifies and uses only the visible edge portions within the read image boundaries to calculate inclination. This segmentation allows accurate measurement despite the document extending beyond boundaries, as it relies solely on the detectable edge segments.
Solution Approach 2:
The patent applies partial action by using only the portion of the document edges that are visible within the read image boundaries for inclination calculation. It doesn't require the entire document to be visible, but rather utilizes the available edge information from the visible portion to accurately determine the document's inclination angle through geometric relationships.
3Device complexity
If conventional methods are used for document inclination detection, then the process is simple when the document is fully visible, but the complexity increases significantly when handling partial visibility cases
Solution Approach 1:
The patent implements a dynamic detection system that automatically selects between two inclination detection methods based on document visibility assessment. A determination unit evaluates whether the document extends beyond boundaries and switches between geometric calculation (for partial visibility) and conventional transform methods (for full visibility). This dynamic approach maintains reliability across different scenarios without requiring manual intervention.
Solution Approach 2:
The patent enables the detection system to self-adapt to different document visibility conditions. The determination unit automatically assesses the visibility state and selects the appropriate detection method without external control. The system serves itself by making intelligent decisions about which algorithm to apply, thereby maintaining both simplicity and reliability automatically.
Data Source
AI summary
A program stored in a non-transitory computer-readable medium causes an image processing apparatus to execute: read image data acquisition processing of acquiring read image data which is generated by optically reading a document and represents a read image including the document; line identification processing of identifying candidate lines which are candidates of lines representing edges of the document through analysis of the read image; and document information identification processing of identifying document information including at least one of a candidate line which represents at least a part of the edges of the document in the read image and an inclination of the document relative to the read image, by using a reference line formed from at least one of the candidate lines identified.