Facial Image Deformation Using External Edge Keypoints
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image processing technologies face issues with holes appearing in facial regions during compression and pixel overlap during stretching, particularly when processing human faces.
Innovation Solution
An image processing method that identifies facial regions, determines keypoint information including external edge keypoints, and uses these to define deformation regions for adaptive deformation processing, avoiding holes and pixel overlap by extending external edge keypoints to facilitate natural image deformation.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Area of moving object
If compressing processing is performed on a facial region, then the facial region size is reduced, but holes appear in the image
Solution Approach 1:
The image processing is divided into multiple deformation regions based on facial keypoint information. Each deformation region is processed independently with appropriate deformation parameters, allowing the facial region to be compressed while maintaining image completeness by coordinating transformations across all segments.
Solution Approach 2:
Multiple deformation regions act as intermediaries between the compression processing and the final image output. These intermediate regions undergo coordinated transformations that prevent holes from appearing while achieving the desired compression effect on the facial region.
2Area of moving object
If stretching processing is performed on the facial region, then the facial region size is increased, but pixels in the image overlap
Solution Approach 1:
The stretching processing is applied to multiple segmented deformation regions rather than the entire facial region at once. Each region is transformed with controlled parameters that prevent pixel overlap while achieving the desired expansion effect.
Solution Approach 2:
Instead of applying uniform stretching to the entire facial region, the processing is applied partially to specific deformation regions with controlled intensity. This partial action approach ensures that pixels are stretched sufficiently to increase facial region size without excessive stretching that would cause overlap.
3Adaptability or versatility
If deformation processing is performed on the facial region, then facial features can be adjusted, but holes or pixel overlap may occur
Solution Approach 1:
The facial region is divided into multiple deformation regions based on keypoint information, allowing different parts of the face to be adjusted independently. This segmentation enables versatile facial feature adjustment while maintaining image quality through coordinated transformations.
Solution Approach 2:
Different deformation parameters are applied to different deformation regions based on their local characteristics and keypoint information. This local quality approach allows each region to be transformed appropriately, achieving versatile facial feature adjustment while preventing holes or pixel overlap in specific areas.
Data Source
AI summary
An image processing method includes: obtaining a first image, identifying a facial region in the first image, and determining keypoint information related to the facial region, where the keypoint information includes keypoint information of the facial region and external edge keypoint information, and a region corresponding to the external edge keypoint information includes the facial region and is larger than the facial region; and determining a plurality of deformation regions based on the keypoint information related to the facial region, and performing image deformation processing on the facial region based on at least part of the plurality of deformation regions to generate a second image.