Gray Scale Image Cleaning for Check Code Line OCR
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Gray scale images used in check code line optical character recognition (OCR) are often polluted by check print backgrounds and video noise, leading to errors and manual labor due to increased contrast between foreground information and noise, making it difficult to produce clean black and white images.
Innovation Solution
A gray scale image cleaning algorithm that generates a normally thresholded black and white image and a skeleton image with altered threshold settings, logically combining them to create a composite image, followed by connected component analysis and recursive thresholding to enhance OCR accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If traditional thresholding is applied to convert gray scale images to black and white images, then the processing speed is maintained, but video noise and background pollution remain in the code line region
Solution Approach 1:
The patent divides the image processing into multiple thresholding passes: a first thresholding operation to create an initial binary image, and a second thresholding operation with different parameters to create a refined binary image. This segmentation of the thresholding process allows noise removal while preserving code line characters, resolving the contradiction between OCR accuracy and processing complexity.
Solution Approach 2:
The patent applies different thresholding parameters and processing operations to different regions of the image, specifically targeting the code line region versus the background. By adjusting threshold values and applying morphological operations locally to the code line area, the system removes noise from specific regions while maintaining overall processing efficiency, thus improving OCR accuracy without uniformly increasing complexity across the entire image.
2Measurement precision
If multiple thresholding operations are applied to remove noise, then OCR accuracy is improved, but processing time and computational complexity increase
Solution Approach 1:
The patent performs preliminary thresholding operations and morphological processing on the entire image before focusing detailed noise removal operations on the code line region. By pre-processing the image to remove obvious noise and simplify structures in advance, the subsequent refined thresholding operations require less computational effort, thus improving code line clarity while minimizing additional processing time.
Solution Approach 2:
The patent applies thresholding operations with parameters that may be more stringent than absolutely necessary for the code line region, but accepts this partial excessive action because it significantly improves noise removal. The system then uses morphological operations to correct any over-processing effects, achieving high code line clarity while managing processing time through selective application of these operations.
Data Source
AI summary
A gray scale image cleaning algorithm for improved check code line OCR. An image processing system for processing a gray scale image is provided that includes: a system for generating a first thresholded black white image from the gray scale image; a system for generating a second thresholded black white image from the gray scale image, wherein the second thresholded black white image is generated with a higher threshold value than the first thresholded black white image; and a system for logically combining the first and second thresholded black white images to generate a composite image.