OCR Pattern Deletion for Logo Detection Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image forming apparatuses face challenges in accurately performing optical character recognition (OCR) on image data, particularly due to erroneous detection of specific patterns like company logos, which affects the accuracy of converting image data into document data.
Innovation Solution
An information processing apparatus with a specific pattern storage unit, pattern comparing unit, character recognizing unit, and file generating unit that detects and deletes specific patterns from image data, performs OCR on the pattern-deleted data, and generates document data, allowing for conversion of specific patterns into different data types (text, image, or combined) to improve accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If OCR processing is performed on image data containing specific patterns like company logos, then the OCR process can be executed on the complete image, but erroneous detection of these patterns occurs reducing accuracy
Solution Approach 1:
The patent extracts and removes specific patterns (such as company logos) from the image data before performing OCR processing. The pattern comparing unit identifies these specific patterns and the image data processing unit removes them, allowing the OCR to focus only on text regions and avoid erroneous detection of logos as text.
2Adaptability or versatility
If all image data is converted to text data, then complete document data is generated, but specific patterns are incorrectly recognized reducing conversion accuracy
Solution Approach 1:
The patent segments image data into different types: text regions and specific pattern regions. The pattern comparing unit identifies specific patterns and separates them from text regions. The file generating unit then converts only the text regions to text data while preserving specific patterns as images, creating adaptively formatted document data with improved recognition accuracy.
Solution Approach 2:
Instead of converting all image data to text data as traditionally done, the patent inverts the approach by identifying and preserving specific patterns as images while converting only the remaining text regions. This inversion prevents erroneous recognition of patterns like logos as text while maintaining document conversion capability.
3Measurement precision
If specific patterns are removed from image data before OCR, then detection accuracy improves, but additional processing steps are required increasing system complexity
Solution Approach 1:
The patent implements a multi-functional pattern comparing unit that performs multiple tasks: identifying specific patterns, determining their positions, and providing this information to the image data processing unit. This universal unit handles both detection and removal functions, improving accuracy while managing system complexity through functional integration.
Data Source
AI summary
An information processing apparatus includes a specific pattern storage unit, a pattern comparing unit, a character recognizing unit, and a file generating unit. The specific pattern storage unit stores specific pattern information in which a specific pattern is registered. The pattern comparing unit compares the image data with a specific pattern that is registered in the specific pattern information by image matching and determines whether or not the specific pattern present in the image data is detected, and when detected, deletes the specific pattern from the image data and outputs pattern-deleted image data. The character recognizing unit performs an OCR process on the pattern-deleted image data, converts a character string in the pattern-deleted image data into text data, and outputs the converted text data as an OCR analysis result. The file generating unit generates document data based on the OCR analysis result.