Neural Network Distortion Correction for Facsimile OCR Accuracy
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Facsimile documents often exhibit textual distortions due to impairments during the transmission process, leading to errors when converted to machine-encoded text using optical character recognition (OCR) engines.
Innovation Solution
A system and method that utilize distortion filter generators to create a training dataset of distorted text documents, which is used to train a neural network model to recognize and correct textual distortions in facsimile images before OCR processing, thereby reducing conversion errors.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If distortion filter generators are used to create training datasets with multiple distortion types, then the accuracy of text recognition in facsimile documents is improved, but the device complexity increases
Solution Approach 1:
The system divides the distortion correction task into multiple independent distortion filter generators, each specialized in generating a specific type of distortion (e.g., compression, elongation, rotation). This segmentation allows the training dataset to be created systematically with various distortion types, improving the neural network's ability to recognize and correct different text distortions in facsimile documents without requiring a monolithic complex system.
Solution Approach 2:
The distortion filter generators create a comprehensive training dataset with pre-distorted text images before the neural network is trained. This preliminary action of generating diverse distorted text samples allows the neural network to learn distortion patterns in advance, enabling accurate recognition and correction of text in actual facsimile documents without requiring complex real-time distortion analysis systems.
2Adaptability or versatility
If multiple distortion filter generators are used to cover various distortion types, then the adaptability of the system to different distortion patterns is improved, but the training data generation process becomes more complex
Solution Approach 1:
Each distortion filter generator is designed to be multi-functional by applying multiple distortion parameters (compression ratios, elongation factors, rotation angles) to the same base text image. This universality allows a single generator to produce various distortion types, expanding the system's adaptability to different distortion patterns while avoiding the need for separate specialized generators for each distortion type.
Solution Approach 2:
The system varies distortion parameters such as compression ratio, elongation factor, and rotation angle across different training samples generated by the distortion filter generators. By systematically changing these parameters, the training dataset covers a wide range of distortion types and severities, enhancing the system's adaptability without requiring a proportional increase in the number of generators or complexity in the data generation process.
Data Source
AI summary
A method includes passing an original text document through distortion filter generators to generate a training dataset that includes distorted text documents. Each distortion filter generator is configured to distort words or letters of words in phrases of text of a facsimile image in a respective unique manner. A neural network model is trained to recognize each respective distortion and match each respective distortion with each respective distortion filter generator based on the training dataset and the original text document. Image data of one facsimile having at least one text distortion is received and inputted to the trained neural network model. The output of the trained neural network model is coupled to an input of an optical character recognition (OCR) engine. The trained neural network model and the OCR engine convert the received image data of the incoming facsimile corrected for the at least one text distortion to machine-encoded text.