Document Image Line Detection Using Multi-Threshold Binarization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting and removing horizontal and vertical lines in document images are often affected by image quality and binarization thresholds, leading to errors in connected component analysis during optical character recognition and document authentication.
Innovation Solution
A method that generates multiple binary images from a grayscale document image using different binarization conditions, detects lines independently in each, and merges the results to effectively remove horizontal and vertical lines by analyzing stroke widths and performing morphological operations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If morphology based line detection is used, then line detection can be performed, but it fails when binarization threshold is improper
Solution Approach 1:
The patent applies multiple binarization thresholds (e.g., 128, 64, 32) to convert the grayscale image into multiple binary images with different threshold values. This allows the morphology-based line detection to be performed on multiple binary representations of the same image, ensuring that at least one binary image will have appropriate threshold characteristics for reliable line detection, thereby resolving the sensitivity to single-threshold impropriety.
Solution Approach 2:
The patent segments the line detection process into multiple independent detections on separate binary images generated from different thresholds. Each binary image is processed independently through morphology operations, and the results are merged. This segmentation allows each detection to operate under its own optimal threshold conditions, overcoming the limitation of single-threshold sensitivity.
2Measurement precision
If Hough transform is used for line detection, then lines can be detected, but detection accuracy is affected by image quality and binarization
Solution Approach 1:
The patent changes the binarization threshold parameter to generate multiple binary images with different threshold values. By doing so, it creates multiple versions of the image that can be processed with Hough transform or other morphology-based methods, ensuring that at least one version will have optimal binarization characteristics for accurate line detection, thereby reducing sensitivity to image quality variations.
Solution Approach 2:
The patent performs preliminary binarization with multiple different thresholds before conducting the line detection. This preliminary action of creating multiple binary representations ensures that the subsequent line detection process has multiple quality-controlled inputs to work from, improving overall detection accuracy while being less sensitive to the quality of any single binarized image.
3Reliability
If run length coding is used for line detection, then lines can be identified, but detection reliability is affected by binarization threshold
Solution Approach 1:
The patent changes the binarization threshold parameter by applying multiple different thresholds (e.g., 128, 64, 32) to the grayscale image, creating multiple binary images. This allows run length coding to be applied to multiple binary representations, ensuring that at least one will have appropriate threshold characteristics for reliable line identification, thereby reducing dependence on any single threshold value.
Solution Approach 2:
The patent segments the detection process into multiple independent run length coding operations on separate binary images. Each binary image is processed independently, and the results are merged. This segmentation ensures that the reliability of line detection is not dependent on a single binarization threshold, as each segment can operate under its own optimal threshold conditions.
Data Source
AI summary
A vertical and horizontal line detection method for document images includes generating multiple binary images from the input grayscale document image based on multiple binarization thresholds, detecting horizontal and vertical lines in each of the multiple binary images independently, and merging the detection results from the multiple binary images. The line detection process for each binary image include applying an opening operation using a vertical or horizontal line as the structuring element, and removing connected components that are not vertical or horizontal lines based on a stroke width analysis. The boundaries of the detected lines are obtained using horizontal and vertical projections.