Spoofing Detection Using Flash Luminance Ratios
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Spoofing detection in face authentication systems is affected by color dependence, as existing methods rely on diffuse reflection models that vary with surface color, leading to inaccurate results for faces with the same three-dimensional shape but different surface colors.
Innovation Solution
A method that calculates a feature value independent of surface color by comparing image frames taken with and without flash, using the ratio of luminance values to determine the authenticity of a face, thereby isolating the three-dimensional shape from color-dependent reflection coefficients.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Device complexity
If diffuse reflection model is used for spoofing detection, then the detection can be performed using simple hardware, but the measurement precision deteriorates due to color dependence
Solution Approach 1:
The patent changes the parameter used for measurement from absolute luminance values (which are color-dependent) to luminance ratios between flash-on and flash-off images. This parameter transformation eliminates color dependence while preserving the ability to detect three-dimensional shape characteristics, thereby improving measurement precision without requiring complex hardware.
Solution Approach 2:
The patent introduces an intermediary calculation step by computing the ratio of luminance values from two different imaging conditions (with flash and without flash). This intermediary ratio serves as a color-independent feature that mediates between the raw color-dependent measurements and the final spoofing determination, enabling accurate detection without direct reliance on absolute luminance values.
2Ease of manufacture
If absolute luminance values are used for feature calculation, then the calculation is simple, but the reliability deteriorates due to color variation
Solution Approach 1:
The patent transforms the parameter from absolute luminance to luminance ratio (flash-on image luminance divided by flash-off image luminance). This parameter change maintains calculation simplicity while eliminating the reliability issue caused by color variation, as the ratio cancels out color-dependent components.
Solution Approach 2:
The patent employs a feedback mechanism by capturing two images under different lighting conditions (with flash and without flash) and using the difference information to correct for color effects. The luminance ratio serves as a feedback-corrected feature that compensates for color variation while maintaining detection reliability.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
This approach enables accurate spoofing detection that is robust against color changes on the face surface, improving the reliability of face authentication by distinguishing real faces from fake ones regardless of surface color variations.
Implementation Method 1
when light is emitted onto the surface of the face of a subject person, the light is diffusely reflected by the surface of the face
Data Source
AI summary
A spoofing detection apparatus 300 includes a face image obtaining unit 301 that obtains a first image frame that includes the face of a subject person obtained with a light-emitting apparatus 320 turned on and a second image frame that includes the face of the subject person obtained with a light-emitting apparatus 320 turned off, a face information extraction unit 302 that extracts first face information, and extracts second face information, a feature value calculation unit 303 that obtains a luminance value of the face portion in the first image frame, obtains a luminance value of the face portion in the second image frame, and calculates, based on these values, a feature value that reflects a three-dimensional shape, and that is independent of colors of a face surface, and a spoofing determination unit 304 that determines authenticity of subject person based on the feature value.