Masked Image Stylization for Preserving Face and Target Region Fidelity
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image stylization technologies fail to maintain a strong association between stylized and original images, particularly in key regions like faces, leading to poor recognition and inconsistency.
Innovation Solution
An image processing method and apparatus that utilizes a stylization processing system, including an encoding model, image reconstruction model, and image stylization model, to process an image and a mask image, ensuring the target region in the stylized image maintains high similarity to the original, using a mask image to distinguish and process the target region with adjusted stylization.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If stylization processing is applied to the entire image, then the image style is transformed effectively, but the association between the stylized image and original image deteriorates, especially in key regions
Solution Approach 1:
The patent applies different processing qualities to different regions of the image. Key regions (identified by mask image) maintain higher fidelity to the original image while non-key regions undergo more aggressive stylization. This is achieved by applying different loss function weights or processing intensities to different spatial regions, resolving the contradiction between overall stylization effectiveness and local content preservation.
Solution Approach 2:
The image is segmented into key regions and non-key regions using a mask image. This segmentation allows the stylization processing system to apply different processing strategies to different segments - preserving important content in key regions while transforming styles in non-key regions, thus maintaining both stylization effectiveness and image association.
2Adaptability or versatility
If strong stylization is applied, then the image style changes significantly, but the recognition of original content deteriorates
Solution Approach 1:
Different stylization intensities are applied to different regions based on their importance. Key regions maintain stronger association with original content for accurate recognition, while non-key regions undergo more significant style transformation. This local differentiation resolves the contradiction between style adaptability and content recognition precision.
3Reliability
If the target region is preserved with high similarity to original, then content recognition is improved, but the overall stylization effect is reduced
Solution Approach 1:
The patent implements local quality control by preserving target regions with high similarity to original images while allowing other regions to undergo full stylization. This is achieved through region-specific loss functions or processing parameters that maintain important content fidelity while achieving overall style transformation, resolving the contradiction between local consistency and global stylization effect.
Solution Approach 2:
By segmenting the image into target regions and non-target regions, the system can apply different processing intensities. Target regions are protected with higher fidelity constraints while non-target regions receive full stylization treatment, achieving both regional consistency and overall stylization effectiveness simultaneously.
Data Source
AI summary
The present disclosure provides an image processing method and apparatus, a storage medium, and an electronic device. The image processing method includes: receiving an image to be processed and a mask image of a target region in the image to be processed; processing the image to be processed and the mask image based on a stylization processing system, to obtain a stylized image associated with the target region; and displaying the stylized image associated with the target region.