Synthesized Image Generation Grouping Adjacent Positions
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for generating synthesized images from group photos often fail to correctly arrange face images of characters with relationships, leading to misplacement of family members, friends, and couples into separate groups when using pre-defined template positions.
Innovation Solution
A method that groups adjacent synthesis positions on a template, determines the order of these groups based on the number of positions, detects relationships between face images, and distributes them accordingly to maintain relationships, such as family or friends, by measuring distance, size, and facial expressions, ensuring related faces are synthesized together.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If face images are pasted into pre-defined template positions without proper ordering, then the synthesis process is simple and fast, but family members, friends, and couples may be misplaced into separate groups
Solution Approach 1:
The patent applies preliminary action by pre-defining template positions and pre-establishing grouping rules before the actual synthesis process. The system pre-processes the image to detect faces and pre-determines their relationships, so that when synthesis occurs, the faces are already organized into groups that should be kept together, maintaining both speed and accuracy
Solution Approach 2:
The patent uses copying by creating a mapping between detected face positions in the source image and corresponding template positions. Instead of directly manipulating the original image, the system copies face information to template positions while preserving relationship data, allowing accurate arrangement without complex real-time processing
2Manufacturing precision
If synthesis positions are grouped by combining adjacent positions, then related characters can be arranged together naturally, but the processing complexity increases
Solution Approach 1:
The patent applies segmentation by dividing the template into multiple groups of adjacent synthesis positions. Each group corresponds to a specific relationship type (e.g., family members, friends, couples). This segmentation allows the system to independently process and arrange different relationship groups, making the complex task of maintaining multiple relationships manageable through modular processing
Solution Approach 2:
The patent implements local quality by assigning different grouping criteria and arrangement rules to different regions of the template. Each local group of synthesis positions has specific characteristics (e.g., family group vs. friend group), and the system applies tailored processing rules to each region, improving overall arrangement accuracy without uniformly increasing complexity across the entire image
Data Source
AI summary
The synthesized image generation method for generating synthesized image data by extracting face images of characters contained in image data and synthesizing the extracted face images with a template into a plurality of synthesis positions which have been set in advance on the template, the method comprising the steps of: forming groups of the synthesis positions by combining the adjacent synthesis positions; determining an order of the groups according to number of synthesis positions belonging to each of the groups; detecting the face images of the characters contained in the image data; detecting a relationship between the detected face images; distributing the face images to the groups in the determined order according to the relationship between the face images; and generating the synthesized image data by synthesizing the distributed face images with the template into the synthesis positions of the groups.