Image Selection via Hierarchical Object Recognition
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing automatic image selection techniques for album creation often fail to provide variety in selected images, leading to monotonous albums, as they do not consider the state or context of the designated object, resulting in a lack of diversity and user satisfaction.
Innovation Solution
An image processing apparatus that recognizes primary and secondary objects in images, calculates scores based on their presence, and selects images for album creation, incorporating features like face detection, object recognition, and scene categorization to ensure a diverse and contextually relevant image selection.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If automatic image selection is performed based on object presence only, then image selection speed is improved, but image variety and user satisfaction deteriorate
Solution Approach 1:
The patent segments the object recognition process into multiple hierarchical levels: primary object detection (e.g., dog), secondary object detection (e.g., ball, bone), and state detection (e.g., running, playing). This segmentation allows the system to evaluate images based on multiple dimensions beyond simple object presence, thereby increasing image variety while maintaining efficient automated selection through structured processing.
2Speed
If only primary object detection is performed, then detection speed is improved, but detection precision deteriorates
Solution Approach 1:
The patent implements preliminary action by first detecting primary objects quickly, then selectively performing secondary object detection and state analysis only on images containing primary objects. This staged approach maintains high detection speed for the initial filter while achieving high precision in state recognition for the final selection, resolving the speed-precision contradiction through sequential processing.
3Measurement precision
If multiple objects and states are recognized, then image selection accuracy is improved, but processing complexity deteriorates
Solution Approach 1:
The patent applies local quality by assigning different levels of analysis to different images based on their content. Images are first screened for primary objects, then secondary objects and states are analyzed only where relevant. This creates a non-uniform processing approach where computational resources are concentrated on critical evaluation points, improving selection accuracy without uniformly increasing processing complexity across all images.
Data Source
AI summary
At least one apparatus recognizes a first object and a second object associated with the first object in a plurality of images, calculates a score for each of the plurality of images based on a result of the recognition of the first object and the second object, and selects an image concerning the first object from the plurality of images based on the score for each of the plurality of images.