Feature Identification in Images via Region Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The increasing resolution of images captured by image capture devices leads to high computational overheads for storage and processing, especially in real-time applications like sporting events or security systems, where complex scenes with many objects require expensive image processing techniques.
Innovation Solution
A method of image processing that involves locating objects in an image, selecting a portion of the image containing the objects, generating a smaller different image with the objects, and detecting a plurality of points corresponding to parts of the objects in this different image, allowing for efficient real-time analysis.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If high resolution images are captured by image capture devices, then image quality is improved, but computational overheads for storage and processing increase
Solution Approach 1:
The patent segments the image processing task into two stages: first identifying candidate regions containing objects of interest, then performing detailed feature analysis only on these segmented regions. This divides the computationally expensive processing from the entire high-resolution image to only relevant portions, reducing overall computational overhead while maintaining analysis accuracy.
Solution Approach 2:
The patent extracts and identifies candidate regions containing objects of interest from the full high-resolution image before performing detailed feature analysis. By separating the candidate region identification step from the detailed analysis step, the system reduces computational overhead by limiting expensive processing to only the extracted regions of interest rather than the entire image.
2Measurement precision
If computationally expensive image processing techniques are performed on high resolution images, then analysis accuracy is improved, but processing time increases
Solution Approach 1:
The patent segments processing into two phases: rapid candidate region identification followed by detailed feature analysis only on identified regions. This segmentation enables the system to maintain high analysis accuracy through detailed processing while reducing overall processing time by applying expensive techniques only to segmented regions of interest rather than the entire image.
Solution Approach 2:
The patent performs preliminary identification of candidate regions containing objects of interest before conducting detailed feature analysis. This preliminary action filters out irrelevant areas, so that subsequent computationally expensive processing is applied only to regions likely to contain relevant features, thereby reducing total processing time while preserving analysis accuracy.
3Measurement precision
If detailed feature analysis is performed on the entire image, then identification accuracy is improved, but computational complexity increases
Solution Approach 1:
The patent segments the image into candidate regions and non-candidate regions, applying detailed feature analysis only to candidate regions. This segmentation maintains identification accuracy for objects of interest while reducing computational complexity by excluding large portions of the image from expensive processing operations.
Solution Approach 2:
The patent extracts candidate regions containing potential objects of interest from the full image before performing detailed feature analysis. By taking out only the relevant regions for detailed processing, the system maintains identification accuracy while significantly reducing computational complexity compared to analyzing the entire image.
4Productivity
If real-time analysis is provided for complex scenes with many objects, then service responsiveness is improved, but processing requirements increase
Solution Approach 1:
The patent segments processing into rapid candidate region identification followed by focused detailed analysis, enabling real-time responsiveness even for complex scenes with many objects. This segmentation reduces processing requirements by limiting detailed analysis to only candidate regions rather than processing the entire scene at full detail.
Solution Approach 2:
The patent performs preliminary identification of candidate regions before detailed analysis, which reduces processing requirements for real-time analysis of complex scenes. By preparing and identifying candidate regions in advance, the system can provide real-time service responsiveness while minimizing the computational burden of detailed processing.
Data Source
Figure 1~2A
Figure 2B~3
Figure 4~5A
AI summary
A method of feature identification in an image is provided, the method comprising the steps of obtaining a plurality of points corresponding to parts of an object from a plurality of images of the object, the plurality of images corresponding to a plurality of different views of the object in a scene, generating at least one three-dimensional model of the object in accordance with the plurality of points corresponding to parts of the object obtained from a plurality of different views, selecting at least a portion of the at least one three-dimensional model of the object as a region of interest and performing feature identification on a portion of at least one of the plurality of images corresponding to the region of interest.