Document Image Classification via Connected Component Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current document classification systems, particularly in high-volume scanning scenarios, face inefficiencies due to the reliance on Optical Character Recognition (OCR) technology, which is slow and incapable of recognizing graphical features, leading to poor accuracy and slow processing times, especially when dealing with documents in different language locales or with similar visual appearances but distinct content.
Innovation Solution
The implementation of an image classification method using connected component analysis to automatically learn features from input images, generating models that classify documents based on visual appearance without the need for OCR, allowing for efficient clustering and classification of documents with similar features, even across different language locales.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If OCR technology is used for document classification, then text recognition capability is improved, but processing speed deteriorates (1-2 pages per second)
Solution Approach 1:
The patent extracts and removes the OCR text recognition step from the document classification process. Instead of using OCR to convert images to text for classification, the system directly processes image features (edges, contours, shapes, colors) to classify documents. This extraction of the bottleneck OCR step enables high-speed classification at 20-30 pages per second while maintaining classification accuracy through visual feature analysis.
2Measurement precision
If OCR technology is used for document classification, then text-based classification is improved, but capability to recognize graphical features deteriorates (cannot recognize logos, shapes, etc.)
Solution Approach 1:
The patent implements a universal image-based classification system that handles multiple document types and features through a single visual analysis framework. The system processes various graphical features (logos, shapes, layouts, colors) and text-containing documents uniformly by extracting visual features directly from images, eliminating the need for separate OCR processing. This multi-functional approach enables recognition of both graphical and text-based documents with the same system.
Solution Approach 2:
The patent replaces the mechanical OCR text recognition system with an image processing-based visual feature analysis system. Instead of converting images to text through OCR, the system directly analyzes visual features (edges, contours, shapes, colors, layouts) from the original images to perform classification. This substitution enables graphical feature recognition while maintaining text-based document classification capability through visual patterns.
3Ease of manufacture
If traditional classifier engines combine computed features of scanned images to generate training data, then training model creation is simplified, but classification accuracy deteriorates
Solution Approach 1:
The patent performs preliminary organization of training data by clustering scanned images into subclasses based on visual similarity before generating training models. This preliminary action groups images with similar visual features (layouts, formats, graphical elements) into distinct clusters, allowing the system to create more accurate and specialized training models for each subclass. This preprocessing step improves classification accuracy by ensuring that training data is more homogeneous and representative of specific document types.
4Measurement precision
If OCR technology is used, then language-specific text processing is improved, but performance across different language locales deteriorates (cannot recognize documents in different languages)
Solution Approach 1:
The patent replaces the OCR text recognition system with an image-based visual feature analysis system that is language-agnostic. Instead of converting text to recognizable characters through OCR (which requires language-specific processing), the system directly analyzes visual features (layouts, graphical elements, text patterns, formatting) from images. This substitution enables the system to classify documents in any language locale without requiring language-specific processing, as it relies on visual patterns rather than text content.
Data Source
AI summary
A method of classifying an input image includes the initial steps of labeling an input image in accordance with a class and extracting at least one connected component from the input image. The method also includes the steps of calculating at least one feature of the input image and generating a model based on the at least one calculated feature. The method also includes the steps of repeating at least one of the previous steps for at least one other input image and comparing the at least one other input image with the model. The at least one other input image is classified in accordance with the class of the model if the at least one calculated feature of the at least one other input image is substantially similar to that of the model.