Image Processor Identifying Important Persons via Relational Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies struggle to accurately determine important persons in image data, often excluding individuals who appear less frequently, leading to unsatisfactory automatic creation of image products like photo books.
Innovation Solution
An image processor that detects face images, classifies them into groups, determines a main person based on frequency, and identifies important persons related to the main person through positional and meta-information analysis, independent of appearance frequency, to prioritize and layout images effectively.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If image layout is based on appearance frequency of persons, then the main person is accurately identified, but important persons who appear less frequently are excluded
Solution Approach 1:
The patent segments the person identification task into two distinct roles: main person determination based on appearance frequency, and important person determination based on relational analysis. This segmentation allows each determination to use appropriate criteria without compromising the other, resolving the contradiction between accurate main person identification and inclusion of important persons who appear less frequently.
Solution Approach 2:
The patent introduces an intermediary mechanism - the determination unit for important persons - that uses detection results (positional relationships, co-occurrence patterns) as intermediate data to identify persons who are important but appear less frequently. This intermediary bridge connects the main person determination with the comprehensive image layout, ensuring no important person is lost.
2Ease of operation
If automatic image product creation is performed, then user convenience is improved, but user satisfaction decreases due to incorrect person prioritization
Solution Approach 1:
The patent implements a dynamic and flexible determination system that can adapt to different scenarios. The important person determination unit can dynamically adjust its analysis based on detection results, allowing the system to maintain high reliability across various family photo book scenarios while preserving the convenience of automatic creation.
Solution Approach 2:
The patent changes the parameters used for person prioritization from solely appearance frequency to include relational parameters such as positional relationships and co-occurrence patterns. This parameter expansion allows the system to accurately identify both main persons and important persons, significantly improving user satisfaction while maintaining automatic creation convenience.
3Device complexity
If only appearance frequency is used for person determination, then processing simplicity is maintained, but important persons are missed
Solution Approach 1:
The patent segments the determination process into two stages: a simple main person determination based on appearance frequency, and an important person determination based on relational analysis. This segmentation maintains processing simplicity for the primary task while adding precision for the secondary task of identifying important persons who appear less frequently.
Solution Approach 2:
The patent applies partial action by using appearance frequency only for main person determination, and excessive action by adding relational analysis specifically for important person determination. This differentiated approach maintains simplicity where sufficient while adding complexity only where necessary, resolving the contradiction between processing simplicity and identification accuracy.
Data Source
Figure 1
Figure 2
Figure 3~4
AI summary
An image processor is provided which is configured to detect face images from a plurality of image data, to perform same person determination processing based on the detected face images to classify the face images into image groups each including face images of a single person, thereby identifying persons corresponding to the respective image groups, to determine at least one person as a main person from among the identified persons, and to determine at least one person highly related to the main person as an important person from among persons except the main person.