Character Region Extraction via Background Inversion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional character extraction methods, such as binarization and color clustering, fail to accurately separate background and characters in complex images with varied colors, leading to reduced character stroke width and potential missed character regions due to aliasing effects.
Innovation Solution
An apparatus and method that generate a candidate character image by creating binarization maps for similar colors, select a background region, and invert the expanded background region to extract the character region, ensuring accurate character extraction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If binarization or color clustering is used to separate background and character, then the separation process is simple, but the separation accuracy deteriorates in complex images with varied colors and external light effects
Solution Approach 1:
The patent divides the image processing into multiple stages: generating multiple binarization maps with different thresholds, selecting candidate background regions, expanding these regions, and inverting to obtain character regions. This segmented approach allows each stage to optimize for its specific task, improving overall separation accuracy while maintaining computational feasibility.
Solution Approach 2:
The patent performs preliminary actions by generating multiple binarization maps with different threshold values before final character extraction. These preliminary maps are used to identify and expand background regions, which are then inverted to obtain accurate character regions. This preliminary processing prevents direct binarization errors from affecting the final result.
2Speed
If conventional binarization is applied to complex images, then processing speed is maintained, but character stroke width is reduced due to aliasing effects
Solution Approach 1:
The patent performs preliminary background region identification and expansion before final character extraction. By pre-processing to identify background areas and expanding them to cover aliasing-affected regions, the method prevents stroke width reduction in the final character region extraction while maintaining efficient processing through algorithmic optimization.
Solution Approach 2:
Instead of directly extracting character regions from the original image, the patent inverts the approach by first identifying and expanding background regions, then inverting the result to obtain character regions. This inversion method effectively counteracts aliasing effects that would otherwise reduce character stroke width, as the expansion process recovers boundary details lost in direct binarization.
3Adaptability or versatility
If color clustering is used to handle various character colors, then color variability is accommodated, but separation accuracy deteriorates when characters are divided by external light effects
Solution Approach 1:
The patent segments the color space processing by generating multiple binarization maps with different threshold values, each handling different color ranges. This segmentation allows the system to process various character colors effectively while the subsequent background expansion and inversion steps resolve issues caused by external light effects dividing character regions.
Solution Approach 2:
The patent changes the threshold parameter across multiple binarization maps to accommodate various character colors. By using different threshold values in different maps and combining them through background region selection and expansion, the system maintains adaptability to color variability while achieving accurate separation even when external light effects are present.
Data Source
AI summary
Provided is an apparatus and method for generating an image for character region extraction. Upon input of an original image, a candidate character image is generated from the original image and a plurality of binarization maps for similar colors are generated from the generated candidate character image. A binarization map including a background region is selected from among the plurality of generated binarization maps, and the background region of the selected binarization map is expanded. The expanded background region is inverted to generate an image including a character region, thereby generating an image for accurately extracting the character region without missing a boundary portion of a character.