Corner Point Correction Using Multi-Region Candidate Filtering
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting corner points in images suffer from low precision due to insufficient information, leading to inaccuracies in image correction, particularly in scanned images distorted by perceptive transformation and stretching-deformation.
Innovation Solution
A method and apparatus that determine initial corner points by identifying first candidate points in a local region and selecting second candidate points based on image information from a larger region, using predetermined conditions to correct the initial corner points, thereby improving precision.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If existing corner point detection methods (e.g., Harris technology) are used, then the detection process is simple and fast, but the detection precision is low due to insufficient information
Solution Approach 1:
The patent divides the image into multiple local regions around each initial corner point, with each region processed independently to identify candidate corner points. This segmentation allows the system to focus computational resources on relevant areas, improving precision without requiring global image analysis, thus balancing complexity and accuracy.
Solution Approach 2:
The patent transitions from analyzing corner points in a single dimension (traditional detection) to analyzing them across multiple dimensions by considering multiple local regions with different sizes and shapes around each initial corner point. This multi-dimensional approach enriches the information available for corner point identification, significantly improving detection precision.
2Measurement precision
If a single local region is used for corner point detection, then the processing is fast and simple, but the precision is insufficient due to limited information
Solution Approach 1:
The patent performs preliminary corner point detection to identify initial corner points first, then uses these as anchors for subsequent multi-region analysis. This preliminary action allows the system to focus subsequent processing only on regions around detected corner points, maintaining efficiency while enabling more comprehensive analysis for improved precision.
Solution Approach 2:
The patent analyzes multiple local regions around each initial corner point, which is more than the single region approach. However, by limiting the analysis to only regions around detected corner points rather than the entire image, the system achieves enhanced precision without excessive processing time, applying partial action strategically.
3Manufacturing precision
If traditional document correction models are used, then the correction process is straightforward, but the correction accuracy is low due to inaccurate corner points
Solution Approach 1:
The patent implements a feedback mechanism where detected corner points are continuously refined through multi-region analysis and candidate selection. The system uses the initial detection results to guide subsequent analysis, and the refined corner points feed back into the correction model, progressively improving correction accuracy while managing complexity through iterative refinement.
Solution Approach 2:
The patent changes key parameters of the correction model by using multiple local regions with different sizes and shapes around corner points, rather than a single fixed region. This parameter change enables the model to adapt to varying local image characteristics, significantly improving correction accuracy without requiring a completely complex new model structure.
Data Source
AI summary
The present invention discloses a method and apparatus for correcting a corner point of an image and an image processing device. The method includes: determining first candidate corner points of an initial corner point in a first local region; obtaining information related to the image in a second local region; selecting, among the first candidate corner points of the initial corner point, the first candidate corner points meeting a predetermined condition, as second candidate corner points of the initial corner point according to the information; and correcting the initial corner point using the second candidate corner points of the initial corner point. The apparatus is configured to perform the processes of the method. The image processing device includes the apparatus for correcting a corner point of an image. With the technology, a roughly detected corner point can be corrected.