Character Recognition from Display Screens via Noise Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional character recognition systems struggle with accurately recognizing characters from high-definition images of computer screens captured by mobile cameras due to image noise, which degrades pixel-level recognition performance and limits their effectiveness.
Innovation Solution
A method and apparatus that analyze the input image to determine its type, apply image effects to distinguish character and background regions, binarize the image, and recognize characters using a combination of frequency domain analysis and classifier-based methods, specifically employing Discrete Cosine Transform (DCT) and image blurring to enhance recognition accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a high definition mobile camera is used to photograph a computer screen, then the camera image resolution is greater than the computer screen resolution, but image noise deteriorates character recognition performance
Solution Approach 1:
The patent extracts and removes image noise through a dedicated noise removal process that separates the noise component from the character information. By identifying and eliminating the harmful noise elements while preserving the character data, the system achieves accurate character recognition despite the high-resolution camera capturing more noise than the original screen resolution
Solution Approach 2:
The patent applies preliminary image processing actions before character recognition, including noise removal and binarization. These preliminary steps prepare the image data by eliminating noise and enhancing contrast, creating an optimized version of the image that is more suitable for subsequent character recognition algorithms to process accurately
2Reliability
If conventional character recognition system is used for camera-based images, then the system is simple to implement, but the system is limited in effectiveness for display screen images
Solution Approach 1:
The patent changes the processing parameters and methodology based on the detected image type. When a display screen image is detected, the system applies specific parameters including noise removal algorithms, binarization techniques, and character recognition parameters optimized for screen images. This parameter adaptation allows the same system to handle both conventional documents and display screen images effectively
Solution Approach 2:
The patent implements dynamic processing by first detecting the image type (document vs. display screen) and then selecting appropriate processing algorithms accordingly. The system dynamically adjusts its behavior based on the input image characteristics, switching between different processing pipelines to maintain high accuracy across different image sources
3Measurement precision
If image sharpening and resolution conversion are applied to computer screen images, then the image resolution is improved, but the operation is not appropriate for general camera-based character recognition systems
Solution Approach 1:
Instead of attempting to recover lost resolution through complex sharpening and upscaling operations, the patent extracts the essential character information directly from the camera image by removing noise and applying binarization. This approach achieves effective character recognition without relying on complex resolution restoration algorithms
Solution Approach 2:
The patent replaces the mechanical/optical approach of image sharpening and resolution conversion with a digital signal processing approach using noise removal algorithms and binarization. This substitution achieves the desired character extraction effect through computational methods rather than attempting to restore the original image quality
Data Source
AI summary
An apparatus and a method for recognizing a character based on an input image is provided. The apparatus includes an input unit configured to receive the input image and a controller configured to select, from the input image, a region of image analysis to be used for image analysis, and to analyze the selected region of image analysis to determine a type of the input image, to apply, to the input image, an image effect for distinguishing a character region and a background region in the input image if the type of the input image indicates that the input image is obtained by photographing a display screen, to binarize output of the image effect according to the determined type of the input image, and to recognize a character from the binarized output of the image effect.