Document Recognition via Color Space Conversion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing document recognition technologies face errors when dealing with color changes during input and delivery, as they primarily rely on accurate character recognition and do not effectively handle color conversions, leading to recognition failures.
Innovation Solution
A method and apparatus for document recognition that utilize color space conversion to generate feature vectors from partial images, allowing for the analysis of document types and extraction of key-value information by comparing font sizes, thicknesses, and positions, and using convolutional neural networks to encode feature maps into binary strings for similarity analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If character recognition is used to check whether characters or words matching desired characters or words exist in a document, then document recognition can be performed, but the method assumes accurate character recognition results and fails when color changes occur during input and delivery
Solution Approach 1:
The patent converts document images from color space to grayscale space, fundamentally changing the parameter of image representation. This transformation makes the recognition system invariant to color changes during input and delivery, resolving the contradiction between maintaining recognition accuracy and adapting to color conversions. The grayscale conversion eliminates color as a variable, allowing consistent feature extraction regardless of original document color or subsequent color changes.
2Device complexity
If only reduced images or partial images are used from original document without color conversion, then processing complexity is reduced, but errors occur in document recognition when the color of the original document is changed into black and white
Solution Approach 1:
The patent applies grayscale conversion to partial images extracted from the document, maintaining low processing complexity while ensuring recognition accuracy under color changes. By converting only the necessary partial regions to grayscale rather than processing entire color documents, the system achieves both simplicity and reliability.
3Reliability
If color space conversion is applied to generate feature vectors from partial images, then the influence of color changes is reduced and recognition accuracy is improved, but additional processing steps are required
Solution Approach 1:
The patent segments the document into partial regions and applies color space conversion only to these segments rather than the entire document. This segmentation approach reduces the overall processing burden while maintaining recognition accuracy, as only critical regions requiring color-invariant processing are transformed.
Solution Approach 2:
The patent performs color space conversion as a preliminary step before feature extraction and recognition. By converting partial images to grayscale beforehand, the system establishes color-invariant features early in the processing pipeline, eliminating the need for complex color handling in subsequent recognition stages.
Data Source
AI summary
An apparatus for document recognition according to an embodiment includes a document type analyzer that analyzes a type of a recognition target document based on document feature vector extracted from one or more partial images obtained by color space conversion of one or more partial regions of the recognition target document, and an information extractor that extracts value information from one or more information search images organized in a grid form based on a position of key information of the recognition target document.