Document Boundary Detection Using Candidate Pixel Classification
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies face errors in calculating the boundary line of document regions due to small brightness differences between documents and backgrounds, and the presence of tabs, which can lead to incorrect detection of boundary pixels and errors in approximating the boundary line.
Innovation Solution
An image processing system that detects candidate pixels for the boundary lines, classifies them into coordinate groups, calculates approximate lines based on these groups, determines a provisional line with the largest number of candidate pixels within a prescribed distance, and uses this line to crop the document region accurately.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If candidate pixels are detected based on brightness difference, then boundary detection is performed, but small brightness differences between document and background cause incorrect detection of boundary pixels
Solution Approach 1:
The patent segments the boundary detection process into multiple stages: initial candidate pixel detection, classification into coordinate groups, and selection of representative pixels. This segmentation allows the system to handle uncertain detections by distributing them into groups and selecting only the most reliable representatives, thereby improving both precision and reliability when brightness differences are small.
Solution Approach 2:
The patent implements a feedback mechanism where detected candidate pixels are classified into coordinate groups, and the distribution of these groups is analyzed to determine representative pixels. This feedback loop allows the system to refine its boundary detection by using the classified data to improve subsequent detection accuracy, especially in cases with small brightness differences.
2Device complexity
If boundary line is approximated using all detected candidate pixels, then boundary line calculation is simplified, but presence of tabs causes errors in boundary line approximation
Solution Approach 1:
The patent applies local quality by classifying candidate pixels into different coordinate groups based on their spatial distribution. Instead of treating all candidate pixels uniformly, the system identifies local patterns and selects representative pixels from each group, thereby maintaining simplicity while improving boundary line precision in the presence of tabs or other local variations.
Solution Approach 2:
The patent performs preliminary classification of candidate pixels into coordinate groups before calculating the boundary line. This preliminary action organizes the data in a way that facilitates accurate boundary line approximation by pre-identifying representative pixels, thus avoiding the need for complex post-processing while maintaining high precision.
3Measurement precision
If coordinate groups are generated by classifying candidate pixels, then representative pixel selection is improved, but processing time increases
Solution Approach 1:
The patent segments the candidate pixels into coordinate groups based on their spatial distribution, which allows for efficient processing by working with smaller subsets of data. This segmentation improves representative pixel detection precision while keeping processing time manageable by avoiding the need to analyze all candidate pixels simultaneously.
Solution Approach 2:
The patent generates multiple coordinate groups and selects representative pixels from each group, which may seem like excessive action, but it actually improves precision by providing redundancy. The system can tolerate the additional processing time because the alternative—incorrect boundary detection—would require even more extensive corrections later.
Data Source
AI summary
Provided are an image processing apparatus, an image processing method, a computer-readable medium storing a computer program and an image processing system improving precision of the calculation of boundary based on candidate pixels constituting the boundary of a side of the document region. The image processing apparatus includes a candidate pixel detector for detecting candidate pixels constituting boundary of sides of the document region, a classifier for classifying coordinates of the candidate pixels respectively detected in plural portions of the boundary into a plurality of groups, an approximate line calculator for calculating a plurality of approximate lines for the boundary based on the coordinates belonging to each of the plurality of groups, and a provisional line determination unit for determining a provisional line for the boundary of the side based on the approximate line having the largest number of candidate pixels within a prescribed distance from the approximate line.