Neural Network Pixel Intensity Adjustment for OCR Contrast
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional techniques for improving optical character recognition (OCR) in document images are ineffective due to noisy backgrounds, lighting variations, and skewing, which confuse OCR algorithms and reduce their effectiveness.
Innovation Solution
The use of deep learning techniques to identify text and non-text pixels in images, adjusting pixel intensities to increase contrast between text and background, thereby reducing abrupt transitions that confuse OCR algorithms.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Object-affected harmful factors
If conventional background removal techniques are used, then background noise is reduced, but abrupt intensity transitions at boundaries are created that confuse OCR algorithms
Solution Approach 1:
The patent applies parameter changes by transitioning from binary background removal (setting pixels to white) to continuous intensity adjustment. The system modifies pixel intensities based on probability maps generated by convolutional neural networks, applying variable intensity changes rather than uniform thresholding. This resolves the contradiction by eliminating abrupt transitions while maintaining background noise reduction.
Solution Approach 2:
The patent replaces conventional mechanical image processing techniques with deep learning-based neural networks. Instead of using traditional thresholding or morphological operations, the system employs convolutional neural networks to generate probability maps and apply intelligent intensity modification. This substitution eliminates the fundamental limitation of conventional methods regarding boundary handling.
2Productivity
If simple thresholding is used to separate text from background, then processing speed is maintained, but recognition accuracy deteriorates due to confused boundaries
Solution Approach 1:
The patent introduces probability maps as an intermediary between the input image and the final modified image. The convolutional neural network first generates probability maps that capture the likelihood of each pixel being text or background, then uses these probability maps to guide intensity modifications. This intermediary enables accurate text-background separation without requiring complex thresholding operations, maintaining both speed and accuracy.
Solution Approach 2:
The patent applies preliminary action by pre-processing the image through convolutional neural networks to generate probability maps before performing the actual intensity modification. This preliminary analysis of pixel characteristics enables subsequent intensity adjustments to be made with high accuracy, improving recognition without significantly increasing processing time.
3Measurement precision
If deep learning convolutional neural networks are applied, then text recognition accuracy is improved, but computational complexity and processing time increase
Solution Approach 1:
The patent applies segmentation by dividing the image processing task into distinct functional stages: (1) generating probability maps through convolutional neural networks, and (2) modifying pixel intensities based on these probability maps. This segmentation allows the computationally intensive neural network to operate only on the original image once, while subsequent intensity modifications are faster and more straightforward, reducing overall computational complexity.
Data Source
AI summary
Systems and methods for image modification to increase contrast between text and non-text pixels within the image. In one embodiment, an original document image is scaled to a predetermined size for processing by a convolutional neural network. The convolutional neural network identifies a probability that each pixel in the scaled is text and generates a heat map of these probabilities. The heat map is then scaled back to the size of the original document image, and the probabilities in the heat map are used to adjust the intensities of the text and non-text pixels. For positive text, intensities of text pixels are reduced and intensities of non-text pixels are increased in order to increase the contrast of the text against the background of the image. Optical character recognition may then be performed on the contrast-adjusted image.