Image Processing Apparatus for High-Luminance Character Extraction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional character recognition systems fail to accurately extract high-luminance characters from scanned business form images due to 'show-through' effects, where characters and images from the back side show through, leading to incorrect Key-Value extraction.
Innovation Solution
An image processing apparatus generates a first binary image and a high-luminance image by adjusting pixel luminance thresholds, and determines whether black pixel clusters in the high-luminance image are extraction targets based on their positional relationship with character images in the binary image, effectively distinguishing between characters and background noise.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If a technique extracts objects with high brightness by binarizing with a threshold, then high-luminance characters can be extracted, but back-side objects that show through are also extracted as false positives
Solution Approach 1:
The patent divides the extraction process into two separate binary images: one for general character extraction and another specifically for high-luminance character extraction. By segmenting the extraction targets and applying different processing logic to each, the system can distinguish between front-side characters and back-side show-through objects.
Solution Approach 2:
The patent introduces a determination process that acts as an intermediary between the two binary images and the final extraction result. This intermediary evaluates the positional relationship between character strings in both images and decides whether to extract high-luminance objects based on their spatial context, preventing false extraction of show-through objects.
2Productivity
If conventional binarization with a single threshold is used, then processing is simple and fast, but high-luminance characters assimilate into the background and cannot be recognized
Solution Approach 1:
The patent segments the binarization process into two parallel operations: standard binarization for general characters and high-luminance binarization for bright characters. This segmentation allows the system to maintain fast processing through efficient binary operations while improving recognition accuracy by ensuring high-luminance characters are properly separated from the background.
Solution Approach 2:
The patent changes the luminance threshold parameter specifically for high-luminance character extraction. By adjusting the threshold to accommodate high-brightness characters, the system can recognize these characters without compromising the processing speed achieved through binary image operations.
3Measurement precision
If high-luminance objects are extracted to improve character recognition, then Key-Value extraction accuracy improves, but positional relationship analysis becomes more complex
Solution Approach 1:
The patent segments the extraction logic into distinct pathways: one for standard characters and another for high-luminance characters. Each pathway has its own extraction rules based on positional relationships, simplifying the overall logic by avoiding the need for a single complex decision-making process.
Solution Approach 2:
The patent performs preliminary binarization and character string identification for both standard and high-luminance objects before the final extraction decision. By preparing the data in advance and organizing it by type, the system reduces the complexity of the final extraction logic and makes positional relationship analysis more straightforward.
Data Source
AI summary
A first binary image is generated by binarizing an input image based on a threshold, a second binary image is generated by changing a pixel that has predetermined high luminance in the input image into a black pixel, and whether a black pixel cluster in the second binary image is made to be an extraction target is determined based on a position of a character image identified based on a black pixel cluster in the first binary image, and a position of the black pixel cluster in the second binary image.