Thin Line Image Resampling With Asymmetric Gaussian Blur
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image resolution conversion methods fail to appropriately handle thin lines when resolution conversion magnification differs between directions, leading to thickness changes or disconnection of lines, especially when using a Gaussian filter with uniform sizes in both directions.
Innovation Solution
An image processing apparatus and method that applies a two-dimensional Gaussian filter with different filter sizes and standard deviations in orthogonal directions to maintain appropriate blurring effects and connect thin lines, even when conversion magnifications differ.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If a Gaussian filter with the same filter size in X and Y directions is applied, then the blurring processing is simple, but thin lines may become disconnected or their thickness may be inappropriately changed when conversion magnification differs between directions
Solution Approach 1:
The patent applies asymmetric filter sizes in the Gaussian filter for X and Y directions when conversion magnifications differ. Specifically, when the conversion magnification in the first direction is larger than in the second direction, the filter size in the first direction is made larger than in the second direction, and vice versa. This asymmetric approach prevents thin lines from becoming disconnected or having inappropriate thickness changes during resolution conversion.
2Reliability
If the filter size is increased to prevent line disconnection, then thin line connectivity is improved, but excessive blurring occurs in the image
Solution Approach 1:
The patent applies different filter sizes in different directions based on the conversion magnification requirements. By making the filter size in each direction proportional to its conversion magnification, the blurring effect is localized appropriately for each direction, preventing both line disconnection and excessive blurring. The filter size in the first direction is set larger than in the second direction when the conversion magnification in the first direction is larger, and vice versa.
3Manufacturing precision
If resolution conversion is performed with different magnifications in X and Y directions, then the resolution is optimized for each direction, but thin lines may be disconnected or their thickness may be inappropriately changed
Solution Approach 1:
The patent resolves this contradiction by applying asymmetric filter sizes that match the asymmetric conversion magnifications. When the conversion magnification in the first direction is larger than in the second direction, the filter size in the first direction is made larger, and vice versa. This asymmetric approach maintains thin line connectivity while allowing different resolution optimization in each direction.
Data Source
AI summary
An image processing apparatus includes a blurring processing unit configured to generate a blurred image by applying a two-dimensional Gaussian filter having a filter size of S1×S2 to an input image, and a resolution conversion unit configured to generate, from the blurred image, a converted image of which a resolution is converted at a first conversion magnification R1 and a second conversion magnification R2. When the blurring processing unit sets a relation of S1>S2, a relation of R2>R1 is established, and when the blurring processing unit sets a relation of S2>S1, a relation of R1>R2 is established. Here, standard deviations of the Gaussian filter are set to be standard deviations σ1 and σ2, respectively. The blurring processing unit applies the Gaussian filter having a filter coefficient of σ2=σ1 to the input image when S2 is different from S1.