2D Facial Image Depth Extraction for Anti-Spoofing
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current anti-spoofing technologies rely on 3D image capture equipment or require user motion to extract depth information from facial images, limiting their effectiveness and accessibility in detecting real versus spoofing entities without these resources.
Innovation Solution
A computer-implemented method using machine learning, specifically convolutional neural networks, to extract depth information from 2D facial images, trained on 3D image data to differentiate between real human and spoofing entities without the need for 3D capture equipment or user motion.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If 3D image capture equipment is used to extract depth information for anti-spoofing, then measurement precision of depth information is improved, but device complexity increases
Solution Approach 1:
The patent creates a virtual 3D representation (depth map) by copying and processing 2D image data through machine learning models. Instead of using physical 3D capture equipment, the system generates synthetic depth information from standard 2D images, effectively copying the functional output of 3D systems without the hardware complexity
Solution Approach 2:
The patent replaces the mechanical/optical 3D imaging system with a computational approach using machine learning models. The mechanical system (3D cameras, structured light projectors) is substituted with software-based depth estimation that processes 2D images algorithmically to produce depth maps
2Device complexity
If user motion is required to extract depth information from 2D images, then device complexity is reduced, but ease of operation deteriorates
Solution Approach 1:
The system performs self-service by automatically extracting depth information from static 2D images without requiring user cooperation or motion. The machine learning model independently processes the image data and generates depth maps, eliminating the need for users to perform specific actions or movements
Solution Approach 2:
The patent performs preliminary processing of 2D images to extract depth information before any authentication decision is made. By pre-computing depth features from the captured image, the system prepares all necessary data in advance, eliminating the need for subsequent user motion or interaction
3Reliability
If 3D image capture equipment is used for anti-spoofing, then reliability of spoofing detection is improved, but ease of manufacture deteriorates
Solution Approach 1:
The patent makes the depth extraction system universal by enabling it to work with standard 2D image capture devices that are already widely manufactured and deployed. The machine learning model can process images from any conventional camera, making the anti-spoofing capability universally applicable without requiring specialized 3D imaging hardware
Solution Approach 2:
The patent changes the fundamental parameter from physical 3D capture to computational 3D reconstruction. By transforming the approach from optical/mechanical parameter space to computational/data space, the system achieves comparable reliability using readily available 2D imaging technology and software processing
Data Source
AI summary
An image processing component is trained to process 2D images of human body parts, in order to extract depth information about the human body parts captured therein. Image processing parameters are learned during the training from a training set of captured 3D training images, each 3D training image of a human body part and captured using 3D image capture equipment and comprising 2D image data and corresponding depth data, by: processing the 2D image data of each 3D training image according to the image processing parameters, so as to compute an image processing output for comparison with the corresponding depth data of that 3D image, and adapting the image processing parameters in order to match the image processing outputs to the corresponding depth data, thereby training the image processing component to extract depth information from 2D images of human body parts.