Nonlinear Image Scaling by Object Distribution Direction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image scaling methods suffer from significant distortion when scaling images with different aspect ratios, particularly when transitioning between standard-definition (4:3) and high-definition (16:9) formats, as they either cause severe distortion, lose main objects, or result in unsatisfactory visual effects due to uneven scaling.
Innovation Solution
An image scaling method that determines the distribution direction of main objects in a source image using face detection, edge detection, or entropy coding methods, and applies a nonlinear scaling approach perpendicular to this direction to minimize distortion, particularly by segmenting and applying different scaling factors to areas of varying importance.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If linear scaling is used to scale images between different aspect ratios, then the scaling process is simple, but severe distortion occurs in the scaled image
Solution Approach 1:
The image is divided into multiple regions based on the distribution direction of main objects. Different scaling methods are applied to different regions: nonlinear scaling is applied to regions containing main objects to preserve their shape, while linear scaling is applied to other regions. This segmentation allows the system to maintain simplicity in processing while reducing distortion in critical areas.
Solution Approach 2:
Different scaling quality levels are applied to different parts of the image. Regions containing main objects (detected through face detection, edge detection, or entropy coding) receive higher quality nonlinear scaling processing, while other regions use standard linear scaling. This local differentiation reduces overall distortion without unnecessarily complicating the entire processing pipeline.
2Manufacturing precision
If edge trimming is used to avoid distortion, then image distortion is reduced, but main objects may be lost
Solution Approach 1:
Before scaling is performed, the system detects and identifies the distribution direction of main objects in the image using face detection, edge detection, or entropy coding methods. This preliminary identification allows the scaling process to be customized for each region, ensuring main objects are preserved while still achieving distortion reduction through directional nonlinear scaling.
3Manufacturing precision
If horizontal nonlinear scaling is used to improve visual effect when main objects are in the middle, then distortion is reduced, but sharp distortion occurs when main objects are on the edge or moving horizontally
Solution Approach 1:
The scaling direction is made dynamic and adaptive rather than fixed. The system automatically determines the distribution direction of main objects and adjusts the scaling direction accordingly. When main objects are detected in the middle, horizontal nonlinear scaling is applied; when they are on edges or moving vertically, the scaling direction is adjusted to vertical nonlinear scaling. This dynamic adaptation ensures consistent distortion reduction across different object positions and movements.
Data Source
Figure 1~2
Figure 3~4
Figure 5~7
AI summary
The present invention relates to image processing technologies, and discloses an image scaling method and apparatus to reduce distortion of a scaled image. The image scaling method includes: determining a distribution direction of main objects in a source image; and scaling the source image to a target image through a nonlinear scaling method according to the distribution direction of the main objects in the source image, where the nonlinear scaling method employs a scaling direction vertical to the distribution direction of the main objects in the source image. The embodiments of the present invention are primarily applicable to the image scaling field.