Adaptive Image Binarization Threshold Map for OCR Text Preservation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Establishing effective brightness thresholds for binarizing grayscale images to prepare them for Optical Character Recognition (OCR) is challenging, as incorrect thresholds can distort images and lose valuable text information.
Innovation Solution
A method that establishes boundaries of image objects, classifies them as suspect or non-suspect, creates a local binarization threshold map for non-suspect objects, and expands it to create a global threshold map for the entire image, ensuring accurate binarization and enhancing OCR effectiveness.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If a single brightness threshold is used for binarization, then the processing is simple and fast, but the text information may be lost and image quality deteriorates
Solution Approach 1:
The patent applies local quality by creating a global binarization threshold map where different regions of the image have different threshold values tailored to their local characteristics. Instead of using a single uniform threshold, the system calculates adaptive thresholds for each pixel based on local image statistics, ensuring that text regions are preserved while background regions are properly binarized, thus preventing text information loss while maintaining processing efficiency.
2Loss of information
If adaptive local thresholds are calculated for each region, then text information is preserved, but the processing complexity and time increase
Solution Approach 1:
The patent applies segmentation by dividing the image processing into distinct stages: first establishing object boundaries, then classifying objects as suspect or non-suspect, creating local threshold maps for non-suspect objects, and finally expanding to a global threshold map. This segmented approach allows adaptive thresholding to be applied efficiently by processing different regions with appropriate methods, preserving text information while reducing overall processing time through strategic division of the binarization task.
Solution Approach 2:
The patent applies preliminary action by first establishing object boundaries and classifying objects before performing the actual threshold-based binarization. This preliminary classification allows the system to identify which regions require careful thresholding (non-suspect objects containing text) and which can be processed more quickly, enabling optimized processing time while ensuring text information is preserved in critical regions.
3Measurement precision
If image objects are classified and processed differently, then binarization accuracy improves, but the algorithm complexity increases
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting the binarization threshold parameter based on local image characteristics and object classification. Instead of using a fixed threshold, the system modifies the threshold value for each region according to local statistics and object type, achieving high binarization accuracy. The complexity is managed by implementing this parameter adaptation through a systematic classification framework that identifies object types and applies appropriate thresholding strategies.
Data Source
AI summary
In one embodiment, the invention provides a method for binarizing an image. The method comprises establishing boundaries of image objects of the image and classifying each image object as either suspect or non-suspect. The method further comprises creating a local binarization threshold map comprising threshold binarization values associated with image objects classified as non-suspect and then expanding the local binarization threshold map to cover the entire image thereby to create a global binarization threshold map for the entire image.