Color Image Upscaling and Multi-Language OCR for Document Processing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current business card reading systems face limitations in OCR accuracy due to low scanner resolutions and are restricted to recognizing cards from a limited number of countries, as they require high-resolution scanning and are not adaptable to diverse languages and alphabets.
Innovation Solution
A method that involves up-scaling color images by separating them into multiple layers, using bi-cubic interpolation for luminance layers and bilinear interpolation for chrominance layers, and implementing a multi-language OCR system that considers international and country-specific field recognition rules to process documents from various countries.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high-resolution scanning is used to improve OCR accuracy, then measurement precision is improved, but scanning time increases and device cost increases
Solution Approach 1:
The patent segments the image into multiple resolution layers (first resolution layer and second resolution layer) and processes them differently. The first layer is processed with high-resolution scanning for accuracy, while the second layer is processed with lower-resolution scanning for speed, allowing selective optimization based on processing needs.
Solution Approach 2:
The patent performs preliminary image acquisition at high resolution and then creates downsampled versions for faster processing. This preliminary high-resolution capture ensures that all necessary detail is available, while subsequent lower-resolution processing steps can proceed quickly without sacrificing accuracy.
2Measurement precision
If high-resolution scanning is used to improve OCR accuracy, then measurement precision is improved, but device cost increases
Solution Approach 1:
The patent divides the image processing into multiple resolution layers, allowing the system to use high-resolution scanning only when necessary for the first layer, while using lower-resolution scanning for subsequent layers, thereby reducing overall device requirements and cost.
Solution Approach 2:
The patent applies high-resolution scanning partially only to the first resolution layer rather than all layers, using lower resolution for the second layer. This partial application of high-resolution scanning maintains sufficient accuracy while reducing the required scanner capability and associated cost.
3Adaptability or versatility
If multi-language OCR processing is implemented to improve adaptability, then versatility is improved, but device complexity increases
Solution Approach 1:
The patent segments the OCR processing into separate modules for different languages and alphabets. Each language module can be independently configured and processed, allowing the system to handle multiple languages without requiring a single complex monolithic OCR engine.
Solution Approach 2:
The patent creates a universal image processing framework that can accommodate multiple languages and alphabets through a common architecture. The same basic processing pipeline can handle different languages by simply changing the recognition parameters and dictionaries, without requiring fundamentally different system components.
Data Source
AI summary
Method for up-scaling a color image prior to performing subsequent processing on said color image, comprising the steps of converting the color image into multiple image layers distinguishable from each other and up-scaling at least one of said multiple image layers. The up-scaling is tuned towards the subsequent processing, for example luminance is upscaled at higher quality than chrominance. Further, a method for interpreting information present on digitally acquired documents, comprising the steps of: (i) determining a country; (ii) identifying a list of languages and character sets in use in said country; (iii) performing optical character recognition simultaneously using all languages and character sets of the list; (iv) performing field parsing to identify fields in the digitally acquired document on the basis of international as well as country-specific field recognition rules; (v) storing the recognized information according to the identified fields in a database.