Face Recognition Pre-Template Encoding with Contrast-Based Compression
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current facial recognition systems that store face templates on storage mediums are not optimal, as they may not be permissible to store and require extensive rework due to format specificity, necessitating a more flexible and independent pre-template solution.
Innovation Solution
A method and apparatus for creating pre-templates by detecting faces, adjusting compression levels based on contrast, encoding, and storing images, which can be later converted to numerical representations for comparison, allowing for flexible storage and use in facial recognition systems without the need for specific template formats.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Adaptability or versatility
If face templates are stored on a storage medium, then face recognition can be performed, but the system lacks flexibility and requires extensive rework when format changes are needed
Solution Approach 1:
The patent segments the face recognition data into two parts: structured face information (stored in database fields) and image data (stored as encoded cropped images). This segmentation allows the structured data to be easily modified without affecting the image storage format, thereby improving system flexibility while reducing rework requirements.
Solution Approach 2:
The patent introduces an intermediary encoding format (encoded cropped images stored in binary or hexadecimal format) that acts as a mediator between the face information and the storage medium. This intermediary format allows the system to store face data without committing to a specific template format, enabling future format changes without extensive rework.
2Reliability
If face templates are stored in a specific format, then recognition accuracy can be maintained, but the format cannot be easily changed
Solution Approach 1:
The patent implements a dynamic storage approach where face information is stored in structured database fields that can be easily modified, while image data is stored in a flexible encoded format. This dynamic structure allows the system to maintain recognition accuracy through proper data organization while enabling format flexibility when needed.
Solution Approach 2:
The patent changes the storage parameters from fixed template formats to a flexible structure consisting of structured face information fields and encoded image data. This parameter change allows the system to maintain accuracy through proper data organization while enabling future format changes without extensive rework.
3Quantity of substance
If compression level is increased, then storage efficiency improves, but image quality may deteriorate
Solution Approach 1:
The patent applies local quality by adjusting compression levels based on the contrast of specific facial regions. High-contrast areas (such as eyes, mouth, nose) are preserved with higher quality and lower compression, while lower-contrast areas use higher compression. This localized quality adjustment maintains recognition accuracy while improving overall storage efficiency.
Data Source
AI summary
Disclosed herein is a method for creating a pre-template for use in facial recognition including detecting a face in a source image, determining face information for the detected face, cropping the source image around the detected face, adjusting a compression level based on a contrast of at least a portion of the detected face, encoding the cropped image based on the adjusted compression level, and storing the encoded cropped image and the face information in a storage medium, wherein the encoded cropped image is subsequently retrieved from the storage medium, decoded and converted to a numerical representation of the detected face to compare the numerical representation with a face in an input image.