AI 3D Face Grid Alignment for Distortion-Free Decorative Overlays
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current image generation systems suffer from fragmentation or distortion when combining decorative elements with faces in images due to reliance on two-dimensional coordinate information for face landmark detection.
Innovation Solution
An image generating system utilizing an image converting module to generate feature points via an AI model, aligning them with a default grid model using a lookup table to create a real-time 3D face model, and combining it with decorative images to enhance precision without distortion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If 2D coordinate information is used for face landmark detection, then the system complexity is reduced, but the face positioning precision deteriorates causing fragmentation or distortion
Solution Approach 1:
The patent transforms 2D face landmark detection into 3D face model generation by introducing depth information through a depth map. The image converting module converts the 2D image into a 3D representation, and the grid model is aligned in this 3D space rather than 2D, thereby improving positioning precision without significantly increasing system complexity
Solution Approach 2:
The patent changes the coordinate system parameters from 2D (x, y) to 3D (x, y, z) by incorporating depth information. The lookup table stores 3D grid point coordinates corresponding to 2D feature points, and the alignment process optimizes 3D transformation parameters including translation, rotation, and scaling to achieve precise face positioning
2Manufacturing precision
If 3D face models are generated with high precision, then the combination accuracy with decorative elements is improved, but the computational resources required increase
Solution Approach 1:
The patent performs preliminary actions by pre-converting the 2D image to a 3D face model with embedded grid points before the actual decorative element combination. The depth map and 3D model are generated in advance, and the lookup table is pre-computed, so that during the combination phase, only alignment transformations are needed rather than full 3D reconstruction
Solution Approach 2:
The patent introduces a lookup table as an intermediary data structure that maps 2D feature point coordinates to 3D grid point coordinates. This lookup table serves as a pre-computed reference that enables efficient alignment without requiring complex real-time 3D calculations during the decorative element combination process
Data Source
AI summary
An image generating system having high face positioning precision includes an image converting module and an image generating module. The image converting module analyzes the face of an image via an artificial intelligence model to generate a plurality of feature points, each with a feature point coordinate. The image generating module saves a lookup table and a default grid model having a plurality of grid points. The number of the feature points is equal to that of the grid points. The lookup table records the grid point coordinate of the grid point corresponding to the feature point coordinate of each feature point. The image generating module finds out the feature points matching the grid points, and aligns at least a portion of the grid points with the feature points corresponding thereto so as to combine the default grid model with the face and generate a real-time 3D face model.