Font Conversion Model for OCR Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing OCR systems face challenges in accurately recognizing characters from images printed with non-recommended fonts, as they cannot correctly collate the character shapes, leading to lower recognition accuracy.
Innovation Solution
An image processing system that acquires a scanned image, extracts character regions, and performs font conversion using a trained model to convert characters from a non-recommended font to a recommended font, enhancing OCR accuracy by generating a converted image suitable for OCR processing.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If OCR is performed directly on images with non-recommended fonts, then the processing speed is maintained, but the recognition accuracy deteriorates
Solution Approach 1:
The patent applies preliminary action by performing font conversion before OCR recognition. The system identifies the font type in the input image, converts it to a recommended font using conversion data, and then performs OCR on the converted image. This preliminary font conversion ensures that the subsequent OCR process achieves high recognition accuracy regardless of the original font used in the document.
Solution Approach 2:
The patent uses an intermediary approach by introducing font conversion data as a mediator between the original image and the OCR process. The conversion data acts as a bridge that transforms non-recommended fonts into recommended fonts, enabling the OCR system to accurately recognize characters from various font types without requiring multiple OCR engines for different fonts.
2Measurement precision
If multiple OCR engines are used to support various fonts, then the recognition accuracy is improved, but the device complexity and processing time increase
Solution Approach 1:
The patent applies universality by creating a unified font conversion approach that works for all font types. Instead of maintaining multiple specialized OCR engines for different fonts, the system uses a single OCR engine combined with a universal font conversion mechanism. The conversion data stores transformation rules that can handle various font types, making the system universally applicable to documents with any font while keeping the processing pipeline simple and fast.
3Measurement precision
If font conversion is performed for all images, then the OCR accuracy is improved, but the processing time increases
Solution Approach 1:
The patent applies local quality by selectively performing font conversion only when necessary. The system first identifies the font type used in the input image, and only performs conversion if the font is not a recommended font. For images already using recommended fonts, the system skips the conversion step and proceeds directly to OCR, thereby maintaining high processing efficiency while ensuring accurate recognition for all font types.
Data Source
AI summary
An image processing system acquires a scanned image obtained by scanning an original, and extracts a character region that includes characters from within the scanned image. The image processing system performs conversion processing, for converting a font of a character included in the extracted character region from a first font to a second font, on the scanned image using a conversion model for which training has been performed in advance so as to convert characters of the first font in an inputted image into characters of the second font and output a converted image. Then, the image processing system executes OCR on the scanned image after the conversion processing.