Character Noise Classification Using Confidence and Distance
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing character recognition methods often incorrectly determine large non-character objects as characters or separate characters as noise, due to reliance on size and distance alone, leading to erroneous classifications.
Innovation Solution
An information processing apparatus that calculates a feature vector for each character, including size, distance from the nearest character, and confidence level, and uses a classification model to determine whether the character is a noise or an actual character, thereby reducing erroneous determinations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If character recognition is performed based only on size and distance from neighboring characters, then the classification process is simple and fast, but erroneous determination occurs where large non-character objects are misidentified as characters or separated characters are wrongly classified as noise
Solution Approach 1:
The patent introduces a confidence level parameter from the character recognition result, in addition to size and distance parameters. This multi-parameter approach transforms the classification from a simple binary decision based on limited parameters to a more nuanced evaluation using multiple parameters including recognition confidence, thereby improving classification accuracy without excessive complexity increase
Solution Approach 2:
The patent employs a classification model as an intermediary component that processes multiple parameters (size, distance, confidence level) and produces a refined classification decision. This intermediary layer separates the raw feature extraction from the final classification decision, allowing complex multi-parameter evaluation while maintaining system modularity and managing overall complexity
2Productivity
If image compression is applied to decrease the number of pixels, then processing speed increases, but recognition accuracy may be affected by loss of character details
Solution Approach 1:
The patent performs character recognition and noise classification on compressed image data before any restoration or expansion operations. By making classification decisions on the compressed data while it still contains the essential character features, the system achieves fast processing without needing to restore full-resolution details first, thus maintaining productivity while ensuring accurate classification
3Adaptability or versatility
If a fixed threshold for size and distance is used to determine noise, then the classification process is straightforward, but it cannot adapt to varying document layouts and character sizes
Solution Approach 1:
The patent replaces fixed static thresholds with a dynamic classification model that adapts its decision boundaries based on the input data characteristics. The model processes size, distance, and confidence level parameters in a flexible manner that automatically adjusts to different document layouts, character fonts, and sizes, thereby achieving high adaptability while the model structure itself remains relatively simple and efficient
Data Source
AI summary
An information processing apparatus includes a processor configured to obtain, for each character of plural characters recognized from an image, (a) position of the character in the image, (b) size of the character, and (c) confidence level of a character recognition result of the character; and determine whether to regard the character as a noise based on a distance between the character and its nearest character, the size of the character, and the confidence level of the character recognition result of the character.