Object Detection Using Saliency and Improper Region Exclusion
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for detecting the main object in an image based on salient regions are ineffective as they fail to accurately distinguish between the main object and non-main objects, leading to erroneous detections that are not useful for applications like auto-focusing and auto-tracking, due to the lack of consideration for the photographer's intention and the inherent difficulty in defining the main object.
Innovation Solution
An object detecting apparatus that identifies candidate regions based on saliency, excludes improper regions set by application-specific conditions, and detects the main object region using a combination of salient region information and improper region exclusion, ensuring the detection result is useful for specific applications.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Speed
If main object detection is performed based on salient region only, then detection speed is improved, but detection accuracy deteriorates due to erroneous detection of non-main objects
Solution Approach 1:
The patent segments the detection process into multiple independent modules: salient region detection unit, candidate region identification unit, application condition setting unit, and detection result evaluation unit. Each module performs a specific function, allowing the system to maintain high detection speed through parallel processing while improving accuracy through multi-stage filtering and evaluation.
Solution Approach 2:
The patent performs preliminary actions by pre-setting application conditions and candidate region criteria before actual detection. The system pre-processes images to identify candidate regions based on saliency thresholds, and pre-defines application-specific conditions that will be used to evaluate detection results, thereby reducing computation time during actual detection while maintaining high accuracy.
2Device complexity
If detection is performed without considering application conditions, then device complexity is reduced, but usefulness of detection result deteriorates
Solution Approach 1:
The patent creates a universal detection framework that can serve multiple applications (auto-focusing, auto-tracking, image processing) through a single integrated system. The apparatus uses common core modules for salient region detection and candidate identification, while allowing application-specific conditions to be configured through software parameters rather than requiring separate hardware systems for each application.
Solution Approach 2:
The patent implements dynamic adaptability by allowing application conditions to be changed based on different use cases. The system can dynamically adjust detection parameters, evaluation criteria, and processing priorities according to the specific application being used, making the system versatile without requiring complete redesign for each application scenario.
3Quantity of substance
If all salient regions are detected as main objects, then detection coverage is improved, but false positive rate increases
Solution Approach 1:
The patent implements feedback mechanisms where detection results are evaluated against pre-set application conditions and candidate region criteria. The system provides feedback by comparing detected regions with expected characteristics (such as region size, position, and application-specific requirements), allowing it to confirm valid detections and reject false positives while maintaining comprehensive detection coverage.
Solution Approach 2:
The patent initially performs excessive detection by identifying all potential candidate regions that meet basic saliency thresholds, then applies selective filtering through application condition evaluation. This approach ensures no potential main object is missed (maintaining high coverage) while systematically eliminating false positives through multiple evaluation stages, including checking whether detected regions satisfy application-specific conditions.
Data Source
AI summary
A detection result of an object useful to an application can be obtained. A designated condition determining unit designates a condition of a region (improper region) which is not used as a detection result of a main object on the basis of application information which was input. An improper region identifying unit identifies an improper region on the basis of the improper region condition which was input. A final region identifying unit identifies a final main object region from a salient main object candidate region output from a candidate region identifying unit and the improper region output from the improper region identifying unit.