Virtual Try-On Image Database Expansion for Eyewear Compatibility
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
The challenge lies in creating a comprehensive and accurate image database for evaluating eyewear compatibility, which is impractical due to the need for thousands of real photographs of diverse users wearing various eyewear combinations, requiring a more efficient method to compile such a database.
Innovation Solution
A method that processes a single user image and a single eyewear image using processing circuitry to define boundaries, merge with a filter image, and overlay onto the user image, allowing for the generation of multiple 'virtual try-on' images, thereby reducing the need for real photographs and expanding the database with diverse combinations using fitting and mounting models.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If real photographs of each desired combination of user and eyewear are acquired, then the image database includes comprehensive and accurate data for fit evaluation, but the database assembly becomes arduous and impractical
Solution Approach 1:
The patent uses image processing to create virtual copies of eyewear on user faces. Instead of photographing each actual combination, the system takes a base user photo and digitally overlays processed eyewear images to generate synthetic try-on images. This copying approach maintains evaluation accuracy while eliminating the need for thousands of physical photographs.
Solution Approach 2:
The patent extracts only the essential components needed for evaluation: the eyewear frame is isolated from its original image by making background pixels transparent, and only the anterior face of the frame is retained. This extraction allows the eyewear to be cleanly overlaid on user images without requiring complete physical photo sessions for each combination.
2Adaptability or versatility
If diverse combinations of users and eyewear are included in the database, then the evaluation covers more facial structures and eyewear types, but the number of required photographs increases to multiples of twenty-five or more
Solution Approach 1:
The patent creates a universal image processing pipeline that can handle any combination of users and eyewear through digital composition. A single base user photograph can be combined with multiple different eyewear images through the overlay process, allowing one photo to serve multiple evaluation purposes across diverse combinations without requiring separate photographs for each pairing.
Solution Approach 2:
The system generates multiple virtual instances of eyewear on the same user face by processing the same base user image with different eyewear overlays. This copying approach allows comprehensive coverage of diverse eyewear types and user combinations while maintaining a small set of actual photographs.
3Manufacturing precision
If traditional photo acquisition methods are used for each eyewear combination, then real-world lighting and positioning are captured, but the process becomes time-consuming and resource-intensive
Solution Approach 1:
The patent performs preliminary processing of eyewear images by creating transparent backgrounds and isolating the anterior face before overlay. This preliminary preparation allows rapid digital composition without requiring time-consuming physical photo sessions. The eyewear images are pre-processed once and then quickly applied to multiple user images through digital overlay operations.
Solution Approach 2:
The patent replaces the mechanical process of physically photographing each eyewear combination with a digital image processing system. Instead of using cameras and physical props for each combination, the system uses computational methods to compose images, substituting digital manipulation for physical photography and dramatically reducing time requirements.
Data Source
AI summary
The present disclosure relates to a method for expanding an image database for evaluation of eyewear compatibility. In particular, the present disclosure relates to a method, comprising receiving a user image, receiving a frame image, processing the received frame image by setting, as transparent, pixels of the received frame image except for an anterior face of the frame, defining, within the processed frame image, a left boundary and a right boundary of the anterior face of the frame, the left boundary and the right boundary corresponding to the left eye and the right eye, respectively, receiving a filter image, processing the received filter image by setting, as transparent, pixels in the received filter image outside the frame based on the left boundary and the right boundary, merging the processed frame image and the processed filter image, and overlaying the merged image onto the received user image.