Virtual Fitting Image Processing with Transparency Adjustment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional virtual fitting technologies struggle to present a natural fitting state of clothes on a user, as pre-superimposed clothing images often appear unnatural when combined with subject images.
Innovation Solution
An image processing device that acquires subject and clothing images using color and depth sensors, adjusts transparency, scaling, deformation, and position of clothing images to generate a second clothing image that naturally fits the subject, using edit values calculated based on body shape parameters and reference position information.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of manufacture
If a pre-acquired clothing image is superimposed on a subject image as it is, then the processing complexity is low, but the fitting state appears unnatural
Solution Approach 1:
The patent applies preliminary action by pre-calculating and storing edit values (transparency, scaling, deformation, position) for clothing images based on reference body shape parameters before actual virtual fitting. When a user tries on clothes, the system retrieves pre-computed edit values matching the user's body type, avoiding real-time complex calculations while ensuring natural fitting appearance. This resolves the contradiction by preparing adjustment parameters in advance, maintaining both processing simplicity and fitting naturalness.
Solution Approach 2:
The patent employs parameter changes by systematically adjusting multiple clothing image parameters (transparency, scaling, deformation, position) based on matched edit values corresponding to the subject's body shape. Instead of using the original clothing image directly, the system transforms the image by modifying these parameters to achieve natural fitting appearance. This resolves the contradiction by using controlled parameter transformations to enhance fitting naturalness without requiring complete re-processing.
2Measurement precision
If multiple clothing images are stored for different body shapes, then the fitting accuracy improves, but the data storage requirements increase
Solution Approach 1:
The patent applies copying by storing only essential edit value parameters (transparency, scaling, deformation, position values) rather than complete clothing images for each body shape. The system creates parameter copies that can be applied to generate fitted clothing images on-demand. This resolves the contradiction by storing compact parameter representations instead of full image data, reducing storage requirements while maintaining fitting accuracy through parameter-based image generation.
Solution Approach 2:
The patent implements universality by creating a standardized edit value data structure that can serve multiple body shapes and clothing types. The same parameter framework (transparency, scaling, deformation, position) is universally applicable across different clothing items and body types. This resolves the contradiction by establishing a multi-functional parameter system that reduces redundant storage while maintaining accurate fitting for various combinations of clothes and body shapes.
Data Source
AI summary
According to one embodiment, an image processing device includes a subject image acquisition module, a first clothing image acquisition module and a second clothing image generator. The subject image acquisition module is configured to acquire subject images which are images of a subject successively picked up by an image pickup module. The first clothing image acquisition module is configured to acquire a first clothing image which is an image of clothes worn by the subject included in the subject images. The second clothing image generator is configured to adjust transparency of a pixel at a predetermined place of pixels constituting the first clothing image, and generate a second clothing image different from the first clothing image.