Polarized Infrared Face Detection via Orthogonal Image Analysis

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Face recognition systems are vulnerable to deception using decoys such as photos or masks, as they cannot reliably distinguish between real and fake faces, leading to potential security breaches.

Innovation Solution

A method utilizing polarized infrared illumination and dual cameras with orthogonal and parallel filters to capture and process images, analyzing the difference between specular and diffuse reflections to extract texture characteristics, and comparing them against a truth model to verify the authenticity of the face, employing wavelet filtering, Fourier transforms, and power-type modeling to differentiate real from fake faces.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Reliability

If conventional face recognition is used, then identification can be performed, but the system is vulnerable to decoys such as photos or masks

Engineering Contradiction:
Improveface verification reliabilityVSAvoiddecoy deception
Core Design Contradiction:
ReliabilityVSObject-affected harmful factors

Solution Approach 1:

The patent uses polarized infrared illumination to detect specular reflections on the face surface. Real skin exhibits specific polarization patterns in specular highlights that differ from photos or masks, allowing the system to distinguish authentic faces from decoys based on optical properties rather than visual appearance alone

Inventive Principle:
Principle #32Color changes

Solution Approach 2:

The system transforms the face verification problem from conventional 2D image matching to analyzing polarization parameters of reflected light. By measuring the polarization state of specular reflections and comparing against truth models, the system detects subtle optical differences that reveal whether a face is real or a decoy

Inventive Principle:
Principle #35Parameter changes

2Reliability

If polarized infrared illumination and dual cameras are used to detect specular reflections, then decoys can be distinguished from real faces, but device complexity increases

Engineering Contradiction:
Improvedecoy detection accuracyVSAvoidillumination and imaging system complexity
Core Design Contradiction:
ReliabilityVSDevice complexity

Solution Approach 1:

The patent divides the face into multiple zones (forehead, nose, cheeks) and analyzes polarization patterns in each region separately. This segmentation allows the system to handle complex polarization data by processing smaller regions independently, comparing each against corresponding truth model zones, thereby managing computational complexity while maintaining high detection accuracy

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The beam splitter acts as an intermediary optical element that directs polarized light from the face to the two cameras with orthogonal and parallel filters. This intermediary component enables the system to capture polarization information efficiently without requiring complex camera positioning or additional moving parts

Inventive Principle:
Principle #24Intermediary (Mediator)

3Measurement precision

If wavelet filtering and Fourier transforms are applied to analyze texture characteristics, then measurement precision of face morphology improves, but processing time increases

Engineering Contradiction:
Improvetexture characteristic precisionVSAvoidimage processing time
Core Design Contradiction:
Measurement precisionVSLoss of time

Solution Approach 1:

The patent pre-computes and stores truth models containing expected polarization and texture characteristics for each face zone during a learning phase. During verification, the system only needs to compare captured polarization patterns against these pre-established models, avoiding the need to perform complex wavelet transforms and Fourier analyses in real-time, thus reducing processing time while maintaining precision

Inventive Principle:
Principle #10Preliminary action

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

Enhances the signal-to-noise ratio of specularities, allowing precise analysis of face morphology and texture, effectively distinguishing real faces from decoys by matching texture characteristics against a database model, thereby improving security by reducing false positives.

Implementation Method 1

at least one illumination means emitting an infrared flux polarized according to a direction of polarization

Methodology Applied
Scientific EffectPolarization: Polarisation

Implementation Method 2

emitting an infrared flux polarized according to a direction of polarization

Methodology Applied
Scientific EffectInfrared radiation: Infrared Radiation

Implementation Method 3

a camera having a polarizing filter orthogonal to the direction of polarization, a camera having a polarizing filter parallel to the direction of polarization

Methodology Applied
Scientific EffectPolarization filtering: Polarisation

Implementation Method 4

a beam splitter placed between the face and the two cameras and intended to divide the luminous flux coming from the face into a luminous flux captured by the camera and a luminous flux captured by the camera

Methodology Applied
Scientific EffectLight splitting: Reflection

Data Source

PatentEP2901370B1Method for detecting a real face
Publication Date: 2016.07.06 MORPHO
  • EP2901370B1 patent drawingFigure 1~2

AI summary

The invention relates to a method for detecting (200) a real face, comprising: - a step for capturing (202) a parallel image of the face captured by the camera having a polarizing filter parallel to a polarization direction and an orthogonal image of the face captured by the camera having a polarizing filter orthogonally to the polarization direction, - a step for computing a difference image resulting from the difference between the parallel image and the orthogonal image, - a step for generating a filtered image by filtering the difference image by wavelet packets on 5 levels, eliminating the low resolution levels, 1 and 5, - a step for dividing the filtered image into three filtered sub-images (forehead, nose, cheeks), - a processing step where the filtered image and each filtered sub-image undergo: o a Fourier transform or a discrete cosine transform, and o modelling during which the decrease profile of the frequencies is modelled in the form of a power-type model (a.x^b+c), - a step for extracting texture characteristics relative to the specularities of the surface of the face from the filtered image and each filtered sub-image, - a step for analysing, for each area of the face, the coefficient 'b' thus calculated and texture characteristics thus extracted relative to the truth model corresponding to said zone of the face, and - a decision making step (216) relative to the authenticity of the face from the result of the analysis step (214).