GAN Image Deformation via Intermediate Pixel Transformation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional image processing methods struggle with large deformations of face images, leading to ghosting issues at face edges and feature positions.
Innovation Solution
An image processing method using a generative adversarial network (GAN) that inputs an original image to generate an intermediate image and pixel transformation information, followed by pixel transformation to produce a target image, effectively handling large deformations.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If traditional image processing methods are used for large deformation, then processing speed is maintained, but ghosting artifacts appear at face edges and feature positions
Solution Approach 1:
The patent introduces an intermediate image as a mediator between the original image and the final deformed image. The generator first produces an intermediate image that captures the deformation intent, then applies pixel transformation to achieve the final large deformation. This two-stage process with an intermediate representation prevents direct large deformation artifacts while maintaining processing efficiency.
Solution Approach 2:
The patent segments the image transformation process into two distinct stages: (1) generating an intermediate image with the generator network, and (2) performing pixel transformation on the intermediate image. This segmentation allows each stage to focus on specific aspects of deformation, improving overall accuracy and reducing ghosting artifacts.
2Adaptability or versatility
If large deformation is applied directly to the original image, then the desired deformation effect is achieved, but the traditional network cannot learn such large deformation resulting in ghosting
Solution Approach 1:
The patent performs preliminary action by first generating an intermediate image that pre-processes the deformation information. The generator network learns to create this intermediate representation that encodes the deformation intent, making the subsequent pixel transformation more accurate and reducing ghosting artifacts in the final large deformation.
3Shape
If the deformation extent is increased to achieve more dramatic effects, then the visual impact is improved, but ghosting artifacts increase due to the traditional network's inability to learn large deformation
Solution Approach 1:
The intermediate image serves as a mediator that decouples the deformation extent from the artifact generation. By transforming the original image to an intermediate representation first, the system can apply larger deformations in the second stage without the traditional network directly learning the complex large deformation mapping, thus reducing ghosting artifacts even at high deformation extents.
Data Source
AI summary
An image processing method and device, an electronic apparatus and a storage medium are disclosed. The image processing method includes: inputting an original image into a generator of a generative adversarial network to obtain an intermediate image and first pixel transformation information; and performing pixel transformation on the intermediate image according to the first pixel transformation information to obtain a target image.