Handwriting OCR Neural Network Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing methods for handwritten characters in scanned documents struggle with accurately separating handwritten characters from printed characters, leading to errors in character recognition due to unclear breaks between items or characters, and often include unwanted printed elements in the recognition process.
Innovation Solution
An image processing apparatus and method using neural networks to extract handwritten character pixels and estimate handwriting areas, allowing for precise separation of handwritten characters and performing OCR processing on the extracted images to improve recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If only handwritten character pixels are extracted from scan images, then handwriting OCR recognition accuracy is improved, but breaks between items or characters become unclear leading to recognition errors
Solution Approach 1:
The patent applies segmentation by dividing the image processing into two distinct components: handwriting pixel extraction and break detection. The break detection unit identifies break positions between handwritten characters and items by analyzing the original scan image, while the handwriting pixel extraction unit separately extracts handwritten character pixels. This segmentation allows the system to maintain clear break information while preserving accurate handwriting character data, resolving the contradiction between improving recognition accuracy and preserving break information.
2Device complexity
If conventional image processing methods are used to separate handwritten characters, then processing simplicity is maintained, but recognition errors occur due to unclear breaks between items or characters
Solution Approach 1:
The patent segments the image processing function into distinct units: a break detection unit that identifies break positions between characters and items, and a handwriting pixel extraction unit that extracts handwritten character pixels. This segmentation adds specialized functionality for break detection while maintaining the overall processing framework, thereby improving recognition reliability without significantly increasing system complexity.
Solution Approach 2:
The break detection unit performs preliminary detection of break positions between handwritten characters and items before the handwriting pixel extraction unit processes the image. By预先 identifying break positions in the original scan image, the system prepares critical separation information in advance, which then guides the subsequent handwriting extraction process to maintain clear character boundaries and improve recognition accuracy.
3Productivity
If handwritten character pixels are extracted using conventional methods, then processing speed is maintained, but breaks between characters are lost reducing OCR accuracy
Solution Approach 1:
The patent segments the processing tasks into parallel operations: break detection that analyzes the original scan image to identify break positions, and handwriting pixel extraction that processes the extracted handwriting regions. This segmentation allows both operations to proceed efficiently, maintaining processing speed while improving break detection accuracy through dedicated analysis of break positions between characters and items.
Data Source
AI summary
An image processing apparatus obtains a read image of a document including a handwritten character, generates a first image formed by pixels of the handwritten character by extracting the pixels of the handwritten character from pixels of the read image using a first learning model for extracting the pixels of the handwritten character, estimates a handwriting area including the handwritten character using a second learning model for estimating the handwriting area, and performs handwriting OCR processing based on the generated first image and the estimated handwriting area.