Image Scaling Error Allocation for Density Uniformity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image scaling methods, such as the nearest neighbor method, result in density unevenness and changes in image form due to pixel insertion or deletion, particularly in fine patterns, leading to noticeable horizontal streaks and local density variations.
Innovation Solution
An image processing apparatus and method that determines pixel operation positions, calculates density differences as errors, and allocates these errors to adjacent pixels to maintain pixel values within an allowable range, preventing isolated points and density variations during scaling.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If the nearest neighbor method is used for image scaling, then image sharpness is maintained, but density unevenness occurs in the form of horizontal streaks in fine patterns
Solution Approach 1:
The patent applies local quality by dispersing pixel insertion positions according to a dispersal pattern (such as V-shaped patterns) rather than inserting pixels uniformly in lines. This causes different regions of the image to experience different insertion behaviors, eliminating the systematic horizontal streaks while preserving local sharpness characteristics.
Solution Approach 2:
The patent introduces asymmetry through the use of non-uniform dispersal patterns (V-shaped arrangements) for pixel insertion. This asymmetric distribution of insertion positions breaks the symmetry of horizontal streaks and creates a more natural, uniform appearance in the scaled image.
2Object-generated harmful factors
If pixel insertion position is dispersed according to a repeating pattern, then density unevenness becomes less noticeable, but local density varies at pixel inserting positions
Solution Approach 1:
The patent applies parameter changes by dynamically adjusting pixel values at insertion positions based on calculated density differences. The pixel value is modified according to the formula: new pixel value = old pixel value + density difference × correction coefficient, where the correction coefficient is determined by the dispersal pattern. This compensates for local density variations while maintaining the overall dispersal pattern benefits.
3Stability of the object's composition
If density of inserted pixels is determined to maintain constant density in the vicinity, then isolated points are generated at pixel inserting positions, resulting in changes of image form
Solution Approach 1:
The patent applies local quality by using different correction coefficients for different regions based on the dispersal pattern. Pixels in different locations receive different levels of correction according to their specific positions in the V-shaped pattern, allowing density compensation without creating uniform isolated points that would distort the overall image form.
Solution Approach 2:
The patent applies parameter changes by introducing position-dependent correction coefficients that vary based on the dispersal pattern. This ensures that density compensation is applied differently at different locations, preventing the creation of isolated points while still maintaining density consistency in the vicinity of insertion positions.
Data Source
AI summary
An image processing apparatus for scaling an image including: an operation position determining section for determining a pixel operation position, which is either a pixel inserting or deleting position, based on a scaling ratio; an inserting/deleting section for inserting/deleting a pixel to/from the pixel operation position determined by the operation position determining section; an error calculating section for calculating a density difference as an error between densities in a vicinity of the inserted/deleted pixel via the inserting/deleting section, before and after the insertion or deletion; and an error allocating section for allocating the error, calculated by the error calculating section, to the inserted pixel and pixels in a vicinity of the inserted or deleted pixel, so that a pixel value, after allocation, of a pixel of an allocation destination, falls within an allowable range estimated from a pixel value of a pixel near the pixel of allocation destination.