Text-Aware Image Restoration Model for Blurred Characters
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies struggle to effectively enhance and restore distorted images, particularly those with blurred or low-resolution text, using artificial intelligence models.
Innovation Solution
An image restoration model is trained using back propagation, combining an encoder to extract feature information, a sub-model to determine text probability maps, and a decoder to generate high-resolution images, with a composite module to integrate textual and non-textual information, enhancing the visibility of characters in distorted images.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing image processing technologies are used to enhance distorted images, then the processing speed is maintained, but the accuracy and visibility of characters in low-resolution images deteriorate
Solution Approach 1:
The model is segmented into distinct functional modules: encoder for feature extraction, composite module for integrating text probability maps with image features, and decoder for reconstruction. This segmentation allows each module to specialize in specific tasks, improving overall character recognition accuracy while maintaining manageable complexity through modular design
Solution Approach 2:
The composite module combines multiple types of information (text probability maps from sub-model and feature information from encoder) into a unified representation. This composite approach integrates complementary information sources to enhance character visibility and recognition accuracy in distorted images
2Reliability
If a simple image restoration model is used, then the device complexity is reduced, but the ability to restore text in distorted images deteriorates
Solution Approach 1:
The model incorporates feedback mechanisms through the composite module that integrates text probability maps with image features, allowing the system to iteratively refine its restoration output. The back propagation training further reinforces this feedback loop by adjusting model parameters based on restoration quality metrics
Solution Approach 2:
The model operates in multiple dimensional spaces simultaneously: processing image pixel data in spatial dimensions while also incorporating text probability maps in character dimension space. This multi-dimensional approach enables comprehensive restoration of both visual and textual information
3Manufacturing precision
If back propagation is performed in all directions including to the sub-model, then the training completeness is improved, but the training time and computational resources increase
Solution Approach 1:
Instead of performing back propagation in all possible directions, the method selectively applies back propagation only to necessary components (encoder and composite module) while excluding the sub-model. This partial action approach maintains sufficient training accuracy for the restoration functionality while significantly reducing computational overhead and training time
Data Source
AI summary
An electronic device may: obtain, from an image, a sub-model trained to output a text probability map indicating one or more characters associated with the image; obtain, using an input image with a first resolution, an output image with a second resolution larger than the first resolution by executing an image restoration model including an encoder to extract feature information from the input image, a composite module to combine the text probability map of the sub-model for the input image and the feature information, and a decoder connected to the composite module; generate information indicating a result of comparison of a ground truth image corresponding to the input image and the output image; and perform training on the image restoration model by performing back propagation based on the generated information along a first direction, out of the first direction and a second direction.