Document Image Entity Extraction via Segmented Noise Removal
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing textual data extraction systems face challenges due to image quality issues in scanned documents, leading to errors in extracted data that require further proofreading.
Innovation Solution
A document processing system that employs a stepwise process to extract entities from documents, utilizing connected components methodology for floating image detection, an ML-based structured data identification model for structured data units, and text spill algorithm for unstructured floating text, thereby improving accuracy and minimizing interference from noise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing OCR-based text extraction techniques are used, then text can be converted into machine-readable form, but extraction accuracy deteriorates due to image quality issues in scanned documents
Solution Approach 1:
The patent segments the document processing into distinct stages: first identifying and removing floating images, then identifying and removing structured data units, and finally performing OCR on the cleaned regions. This segmentation allows each processing stage to focus on specific task requirements, improving overall extraction accuracy by preventing noise from interfering with text recognition.
Solution Approach 2:
The patent performs preliminary actions by identifying and removing floating images and structured data units before conducting OCR text extraction. This preliminary cleanup of the document image removes potential sources of error and interference, ensuring that the subsequent text extraction operates on clean, high-quality regions.
2Productivity
If traditional OCR processes the entire document, then all text regions are processed, but noise from floating images and structured data reduces extraction quality
Solution Approach 1:
The patent extracts and removes floating images and structured data units from the document image before performing OCR. By taking out these non-text elements, the system prevents them from being misrecognized as text, thereby maintaining high extraction accuracy while preserving productivity through efficient region-based processing.
Solution Approach 2:
The patent introduces an intermediary processing stage between document input and OCR that identifies and removes floating images and structured data units. This intermediary layer acts as a mediator that cleans the document image, ensuring that only valid text regions are passed to the OCR engine for accurate extraction.
Data Source
AI summary
A document processing system processes a document image to identify document image regions including floating images, structured data units, and unstructured floating text. A first masked image is generated by deleting any floating images from the document image and a second masked image is generated by deleting any structured data units from the first masked image. The structured data units and the unstructured floating text are thus identified serially one after another. Textual data is extracted from the structured data units and the unstructured floating text by processing the corresponding document image regions via optical character recognition (OCR). Entities are extracted from the textual data using natural language processing (NLP) techniques.