Character Detection in Overlapped Translucent Images
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional methods fail to accurately detect characters overlapped by transparent images, as the entire overlapped region is incorrectly classified as a text region.
Innovation Solution
A character detection apparatus that calculates pixel frequencies in overlapping regions and generates replacement images based on these frequencies to distinguish characters from the background, using techniques like pixel replacement and closing processing to enhance detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional character detection methods are used, then the detection process is simple, but characters overlapped by transparent images cannot be detected accurately
Solution Approach 1:
The overlapping region is divided into multiple blocks, and each block is further divided into multiple regions. Character detection is performed on each region separately using frequency analysis of pixel gradations. This segmentation allows the system to distinguish characters from transparent image portions by analyzing local pixel distribution patterns rather than treating the entire overlapping region as a single text block.
Solution Approach 2:
Different regions within the overlapping area are analyzed with different characteristics. The system calculates frequency of appearance for pixels at each gradation level locally within each region, allowing it to identify regions with character-like pixel distributions versus regions with uniform transparent image characteristics. This local analysis enables accurate character detection even when characters are overlapped by transparent images.
2Measurement precision
If the entire overlapping region is classified as text region, then the classification process is simple, but character detection precision deteriorates
Solution Approach 1:
The overlapping region is segmented into multiple blocks and regions, allowing differential analysis of each sub-region. Instead of applying a uniform classification to the entire overlapping region, the system analyzes each region's pixel frequency characteristics independently, enabling precise identification of characters versus transparent image portions.
Solution Approach 2:
The system changes the analysis parameter from binary classification (text or non-text) to multi-level frequency analysis. By calculating the frequency of appearance for pixels at each gradation level within each region, the system can distinguish characters from transparent images based on their unique pixel distribution patterns, improving detection precision while maintaining processing simplicity.
Data Source
AI summary
A character detection apparatus is provided that detects, from an image including a first image representing a character and a second image representing a translucent object, the character. The character detection apparatus includes a calculating portion that, for each of blocks obtained by dividing an overlapping region in which the first image is overlapped by the second image, calculates a frequency of appearance of pixels for each of gradations of a property, and a detection portion that detects the character from the overlapping region based on the frequency for each of the gradations.