Binarization Method for Scanned Image OCR Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional character recognition techniques struggle with binarization of scanned images containing areas with varying luminance, leading to inaccurate OCR processing, as they use a single threshold value for the entire image, which is insufficient for handling high and low luminance character areas simultaneously.
Innovation Solution
An image processing apparatus that performs area division processing using a first binarization method for quick preview display and switches to a higher accuracy second binarization method for OCR processing, determining the suitability of the binary image for OCR based on the extracted text areas and adjusting the binarization method as needed.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If a single threshold value binarization method is used for the entire image, then processing speed is improved for quick preview display, but character recognition accuracy deteriorates when the image contains areas with varying luminance
Solution Approach 1:
The patent divides the scanned image into multiple regions based on luminance characteristics, applying different binarization threshold values to different regions. This segmentation allows the system to maintain fast processing through automated region-based analysis while achieving accurate character recognition in each specific luminance zone, directly resolving the contradiction between speed and accuracy.
Solution Approach 2:
The patent implements local quality by determining appropriate binarization threshold values for each region based on its specific luminance characteristics. Instead of using a uniform threshold for the entire image, the system adapts the threshold locally to match each region's lighting conditions, thereby maintaining high character recognition accuracy across varying luminance areas while preserving processing efficiency.
2Loss of time
If a high-speed binarization method is used for preview display, then processing time is reduced, but image quality deteriorates for OCR processing
Solution Approach 1:
The patent segments the image processing into two distinct phases: a quick preview phase using fast binarization for immediate display, and a subsequent OCR phase using refined region-based binarization for high-quality character recognition. This segmentation allows the system to minimize preview preparation time while ensuring high binary image quality for OCR processing.
Solution Approach 2:
The patent performs preliminary region division and threshold determination based on luminance characteristics before OCR processing. By pre-analyzing the image structure and determining optimal binarization parameters for each region in advance, the system prepares high-quality binary images for OCR without compromising preview display speed, as the preliminary analysis is efficiently automated.
3Device complexity
If one threshold value is determined for the entire input image, then device complexity is reduced, but adaptability to different luminance conditions deteriorates
Solution Approach 1:
The patent automatically segments the image into multiple luminance-based regions and determines appropriate threshold values for each region. This segmentation approach enhances adaptability to varying luminance conditions while maintaining relatively simple device complexity, as the region division and threshold selection are performed through automated algorithms rather than complex manual configuration.
Solution Approach 2:
The patent dynamically changes binarization threshold parameters based on the luminance characteristics of different image regions. By automatically adjusting the threshold parameter to match each region's lighting conditions, the system achieves high adaptability to varying luminance environments while keeping the overall system complexity manageable through algorithmic parameter adaptation.
Data Source
AI summary
An image processing apparatus including: an analysis unit configured to extract a text area by performing area division processing for a binary image obtained by binarizing the scanned image by a first binarization method; a determination unit configured to determine a binary image used in OCR processing; and a character recognition unit configured to perform the OCR processing by using the binary image determined by the determination unit for the text area extracted by the analysis unit, and the determination unit: in a case where a binary image used in the area division processing is suitable to the OCR processing, determines the binary image as a binary image used in the OCR processing; and in a case where a binary image used in the area division processing is not suitable to the OCR processing, generates a binary image by a second binarization method whose accuracy is higher than that of the first binarization method and determines the generated binary image as a binary image used in the OCR processing.