Document Image Quality Scoring for More Reliable OCR Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing OCR systems struggle to accurately determine the quality of documents, leading to inconsistent and erroneous data extraction due to variations in document quality, which current methods fail to address.
Innovation Solution
A system and method for determining document quality using a Document Quality Metric (DQM) determination device that segments document images into regions, employs multiple OCR modules and NLP-based text matching, segregates datasets into good, medium, and bad classes, and trains machine learning models to classify image quality, ultimately calculating a cumulative quality score.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If OCR systems are used to extract data from documents, then data extraction capability is improved, but accuracy deteriorates when documents are of poor quality
Solution Approach 1:
The system performs quality assessment of document regions before applying OCR extraction. By evaluating image quality metrics (sharpness, noise, illumination) in advance, the system identifies suitable regions for OCR processing, ensuring that extraction is only attempted on regions meeting quality thresholds, thereby maintaining high accuracy while preserving productivity
Solution Approach 2:
The system divides the document into multiple regions and assesses quality independently for each region. Different regions can have different quality levels, and the system selectively applies OCR only to regions with sufficient quality, allowing high-quality regions to contribute to productivity while low-quality regions are excluded to maintain accuracy
2Measurement precision
If multiple OCR algorithms are used to improve accuracy, then consistency deteriorates due to varying results across algorithms
Solution Approach 1:
The system implements a feedback mechanism where quality assessment results guide the OCR processing decision. The quality metrics provide feedback on region suitability, and this feedback is used to determine whether to apply OCR and which algorithm to use, ensuring consistent decision-making criteria across different documents and regions
3Measurement precision
If quality assessment is performed before OCR, then extraction accuracy is improved, but processing time increases
Solution Approach 1:
The system segments the document into multiple regions and performs quality assessment on each segment independently. This allows parallel processing of quality evaluation across regions, reducing the total time required compared to assessing the entire document sequentially, while still maintaining accurate quality-based selection for OCR processing
Data Source
Figure 1
Figure 2
Figure 3A
AI summary
A method and system of determining quality of a document image is disclosed that includes segmenting, by one or more processors, a document image into a plurality of regions each of which comprises text data. The plurality of regions is classified into one of a plurality of image quality classes based on a determination of a highest prediction value from one of a plurality of machine learning models. The plurality of machine learning models is trained corresponding to one of the plurality of image quality classes. A cumulative quality score for the image is computed based on a weighted average of a number of regions classified into each of the plurality of image quality classes. The quality of the image is determined based on the cumulative quality score.