Face Recognition Using Discrete Transform Encoded Local Binary Patterns
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current face recognition technologies using local binary patterns (LBP) in partial-face images are computationally expensive and inefficient, failing to provide accurate and cost-effective solutions for near real-time applications like surveillance and access control.
Innovation Solution
A method involving discrete transforms to generate transform coefficients, followed by LBP encoding and conversion to decimal numbers, which are then used for feature matching in image data, enhancing the accuracy and efficiency of face recognition.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If local binary patterns are applied to raw pixel intensities in partial-face images, then face recognition can be performed, but the computational expense increases and accuracy decreases
Solution Approach 1:
The patent applies a discrete transform (such as DCT or DFT) to the image data before extracting local binary patterns. This preliminary transformation converts the image into frequency domain coefficients, which then serve as the input for LBP extraction. This approach prepares the data in advance to reduce computational complexity during the actual recognition process while maintaining accuracy.
Solution Approach 2:
The patent changes the parameter space from raw pixel intensities to transform coefficients (frequency domain representation). By working with transformed coefficients rather than original pixel values, the system achieves more efficient computation and better discrimination power for partial-face recognition, resolving the contradiction between accuracy and computational efficiency.
2Measurement precision
If exhaustive comparison methods are used to extract unique information from partial-face images, then recognition accuracy improves, but computational cost increases
Solution Approach 1:
The system performs preliminary discrete transform on the image data to create a compressed representation in the frequency domain. This pre-processing step reduces the amount of data that needs to be compared exhaustively, allowing accurate recognition with lower computational cost by working with transformed coefficients rather than raw pixel data.
Solution Approach 2:
The patent extracts local binary patterns from the transform coefficients rather than from raw pixel intensities. This extraction approach isolates the most discriminative features in a compressed representation space, enabling exhaustive comparison to be performed more efficiently with reduced computational energy while maintaining high recognition accuracy.
Data Source
AI summary
Determining a match between the subjects of first and second images as a function of decimal-number representations of regions of the first and second images. The decimal-number representations are generated by performing discrete transforms on the regions so as to obtain discrete-transform coefficients, performing local-bit-pattern encoding of the coefficients to create data streams, and converting the data streams to decimal numbers. In one embodiment, the first and second images depict periocular facial regions, and the disclosed techniques can be used for face recognition, even where a small portion of a person's face is captured in an image. Subspace modeling may be used to improve accuracy.