Facial Image Generation Using Feature-Block Inheritance Networks
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing face image generation solutions using general-purpose processing networks produce composite images that lack harmony and naturalness, making them difficult to believe as real images, and require costly databases for training.
Innovation Solution
A specialized inheritance network and attribute enhancement network are used to generate composite face images that inherit facial features from two input images, allowing precise control over attributes like age and gender, trained without a database of familial relationships using existing face databases.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional image processing methods are used for face beautification, then processing speed is fast, but the generation quality and realism of facial images are poor
Solution Approach 1:
The method performs preliminary actions by pre-processing the input image to extract facial features, landmarks, and structural information before the main generation process. This preparation work is done in advance to enable faster and more accurate face generation and beautification processing in subsequent steps.
Solution Approach 2:
The patent introduces intermediate representations including facial landmark points, structural information, and feature embeddings that act as mediators between the input image and the final generated facial image. These intermediaries enable the model to capture essential facial characteristics while maintaining processing efficiency.
2Manufacturing precision
If detailed facial feature processing is performed to improve generation quality, then image realism improves, but computational complexity increases
Solution Approach 1:
The method segments the facial image processing into multiple independent components including landmark detection, structural feature extraction, texture analysis, and generation stages. Each component processes specific facial elements separately, which improves overall image realism while managing computational complexity through modular processing.
Solution Approach 2:
The patent applies local quality by processing different facial regions with appropriate levels of detail and complexity. Critical features like eyes, lips, and nose receive more detailed processing to enhance realism, while less critical areas use simpler processing, thereby balancing image quality with computational efficiency.
3Measurement precision
If comprehensive facial structure analysis is conducted to maintain structural consistency, then generation accuracy improves, but processing efficiency decreases
Solution Approach 1:
The system performs preliminary facial structure analysis by detecting landmarks and extracting structural features before the main generation process. This pre-processing establishes an accurate structural framework that guides subsequent generation steps, ensuring structural consistency while improving overall processing efficiency.
Solution Approach 2:
The patent implements feedback mechanisms where the generated facial image is continuously compared against the extracted structural information and landmarks. This feedback loop ensures that the generated image maintains structural consistency with the input image while allowing for efficient optimization during the generation process.
Data Source
Figure 1
Figure 2
Figure 3
AI summary
Disclosed are a data processing method and device for facial image generation, a medium, and a computer device. The data processing method comprises: obtaining a first facial image (IMA) and a second facial image (IFA); obtaining M first image blocks corresponding to facial features in the first facial image (IMA), and obtaining N second image blocks corresponding to facial features in the second facial image (IFA); transforming the M first image blocks and the N second image blocks into a feature space to generate M first feature blocks and N second feature blocks; selecting some of the first feature blocks and some of the second feature blocks according to a specific control vector; generating a first synthetic feature map on the basis of at least the selected some of the first feature block and some of the second feature block; and inversely transforming the first synthetic feature map back to an image space to generate a third face image, wherein M and N are natural numbers.