Image Segmentation Importance Calculation for Precise Area Trimming
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional image processing technologies face challenges in accurately identifying and isolating important areas in images that include both people and objects, leading to difficulties in cutting out relevant sections, especially when both are present in the same image.
Innovation Solution
An image processing device that divides images into segments and calculates the importance of each segment based on relationships with other segments within the same image or across multiple images, allowing for precise identification and isolation of important areas without relying solely on object recognition technology.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If object recognition technology is used to identify important areas in images, then face recognition precision is maintained at a predetermined level, but object recognition precision for generic objects is insufficient and unreliable
Solution Approach 1:
The image is divided into multiple segments, and importance degrees are calculated for each segment based on different criteria (face recognition results, object recognition results, and segment relationships). This allows the system to leverage the strengths of different recognition methods for different parts of the image while maintaining overall reliability.
Solution Approach 2:
The patent combines multiple recognition approaches (face recognition and object recognition) and multiple evaluation criteria (within-image segment relationships and across-image segment relationships) to calculate a comprehensive segment importance degree. This merging approach compensates for the weaknesses of individual methods and achieves both precision and reliability.
2Device complexity
If only object recognition technology is used to cut out areas from images, then the process is simplified, but the ability to accurately identify important areas containing both people and objects is insufficient
Solution Approach 1:
The image processing is segmented into distinct stages: dividing the image into multiple segments, calculating importance degrees for each segment using multiple criteria, and finally selecting segments based on importance thresholds. This segmentation allows complex multi-criteria evaluation to be performed in a structured, manageable way.
Solution Approach 2:
The segment importance calculation mechanism serves multiple functions: it evaluates face-containing segments, object-containing segments, and segments important for contextual relationships simultaneously. This multi-functional approach maintains processing simplicity while achieving comprehensive important area identification.
3Adaptability or versatility
If conventional face recognition and object recognition technologies are used separately, then each technology can operate independently, but the ability to cut out areas containing both family members and objects like towers is difficult
Solution Approach 1:
The patent merges face recognition results, object recognition results, and segment relationship analysis into a unified segment importance evaluation framework. This allows the independently operating recognition technologies to work together synergistically, identifying areas containing both people and objects efficiently.
Solution Approach 2:
Segment relationships serve as an intermediary mechanism that connects face recognition and object recognition results. By evaluating how segments relate to each other within and across images, the system can identify important areas containing both family members and objects even when they appear in different locations or contexts.
Data Source
AI summary
Segments included in an image I are each classified as one of object (i.e., person) segments OS1 and OS2 and foreground segments FS1 and FS2. With respect to each of the foreground segments FS1 and FS2, an importance degree is calculated based on a composition of the image I and relations between the foreground segment of the image I and a foreground segment of an image other than the image I.