Document Image Binarization Using Pixel-Wise Local Thresholds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing systems fail to accurately distinguish between the backgrounds and characters in the binarized image of a document such as a security, leading to lower character recognition accuracy due to interference from background elements.
Innovation Solution
An image reading apparatus and method that sets individual thresholds for background and character pixels based on brightness values of neighboring pixels, using a control section to generate a binarized image by setting thresholds for each pixel, enhancing the distinction between background and character pixels.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a single threshold is used for binarization processing of the entire read image, then the processing is simple and fast, but the character recognition accuracy deteriorates due to interference from background elements
Solution Approach 1:
The patent applies local quality by setting different threshold values for different regions of the image. Specifically, it calculates separate thresholds for character pixels and background pixels based on their respective brightness value distributions, allowing each region to be binarized with an optimal threshold tailored to its local characteristics, thereby improving character recognition accuracy while maintaining processing efficiency.
2Measurement precision
If individual thresholds are set for each pixel based on n×n brightness values, then the distinction between background and character pixels is improved, but the processing time and computational load increase
Solution Approach 1:
The patent segments the image processing into distinct steps: first separating pixels into character and background groups, then calculating thresholds for each group independently based on their brightness value distributions. This segmentation allows efficient processing by avoiding the need to analyze all n×n brightness values for every single pixel, thereby reducing computational load while maintaining accurate pixel distinction.
Solution Approach 2:
The patent applies partial action by calculating thresholds based on representative brightness value distributions of character and background regions rather than performing exhaustive analysis on every pixel's complete n×n neighborhood. This approach achieves sufficient discrimination accuracy without the excessive computational cost of full local analysis for each pixel.
Data Source
AI summary
The image reading apparatus includes a reading section and a control section. The control section includes an image acquisition unit, a threshold setting unit, and a binarization processing unit. The threshold setting unit performs threshold setting processing including second processing and third processing, thereby setting a threshold used for binarization processing for each pixel of a read image. The second processing sets a threshold for binarizing a background pixel as a pixel of a background image into white. The third processing sets a threshold for binarizing a character pixel as a pixel of a character image into black. The binarization processing unit generates a binarized image from the read image based on the threshold set for each pixel. The threshold setting unit sets a threshold determined based on each brightness value of n×n pixels containing a pixel as a pixel of interest for the pixel of interest.