Edge Detection Using Signed Difference Values to Prevent Line Broadening
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing edge detection methods for character recognition in color and gray scale documents often result in thick character outlines, inside voids, and broadened line widths, leading to character overlap and reduced image readability.
Innovation Solution
An image processing apparatus and method that converts image information into suitable formats for edge detection, calculates difference values between adjacent pixels, and adds the absolute value of negative differences to generate edge images, preventing inside voids and broadening of character lines by adjusting pixel values accordingly.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If edge detection is performed using absolute values of difference values of density, color and the like between adjacent pixels, then edge images can be obtained from documents, but inside voids are caused with the part along the outline of the character being thick and inside thereof being light, and line width of the character broadens
Solution Approach 1:
The patent changes the parameter used in edge detection by introducing a directionality component. Instead of using only absolute difference values, the invention uses difference values with directional information (positive or negative signs), and selectively applies absolute values only when necessary. This parameter change allows the system to maintain edge detection accuracy while preventing the broadening of character lines and creation of inside voids.
Solution Approach 2:
The patent inverts the conventional approach by not always applying absolute values to difference values. Instead, it preserves the sign information of difference values and only converts to absolute value when the difference value is negative and needs to be adjusted. This inversion of the standard edge detection procedure eliminates the harmful effects of line broadening while maintaining edge detection capability.
2Measurement precision
If conventional edge detection methods are used, then character outlines can be separated from background, but characters contact each other and image readability is reduced
Solution Approach 1:
By changing the parameter from absolute difference values to signed difference values with directional information, the patent achieves more precise character separation without excessive line thickness. This parameter change prevents characters from contacting each other while maintaining clear separation from the background, thereby improving image readability.
3Ease of manufacture
If absolute values of difference values are used in edge detection, then edge images are generated, but line width of characters broadens
Solution Approach 1:
The patent modifies the parameter application by selectively using absolute values only when difference values are negative and need adjustment. This selective parameter change maintains the ease of edge image generation while precisely controlling line width, preventing unnecessary broadening that would occur with universal absolute value application.
Data Source
AI summary
Converting inputted information of an image into information suitable for the edge detection, and calculating difference value difference value of adjacent pixels of the image, and adding an absolute value of the difference value to the calculated difference value of the pixels corresponding to the adjacent pixels if the difference value is negative, to generate an edge image.