Face Recognition Using Frequency Domain Color Analysis
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional human face recognition technologies require high-definition images to accurately determine if a face is from a human or a manikin, leading to increased false rates with low-definition images.
Innovation Solution
A human face recognition method and apparatus that processes red, green, and blue component statistic information using independent component analysis, transforming the derived information into frequency domains to calculate energy values within specific ranges, allowing for accurate determination of human presence regardless of image quality.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional human face recognition technologies are used to determine whether eyes of a human being are shown in human face images, then accurate determination can be achieved, but high-definition images must be used which increases the false rate with low-definition images
Solution Approach 1:
The patent changes the parameter space from spatial domain (pixel-level eye detection requiring high definition) to frequency domain (spectral analysis of color components). By transforming color statistic information into frequency domain and analyzing energy distribution, the system can detect human characteristics without relying on high spatial resolution, thus resolving the contradiction between measurement precision and reliability across different image qualities
Solution Approach 2:
The patent introduces frequency domain analysis as an intermediary between the input images and the recognition decision. Instead of directly analyzing eye structures in spatial domain, the system uses frequency spectral information as a mediator to indirectly detect human characteristics, enabling accurate recognition without high-definition images
2Measurement precision
If high-definition images are used to ensure accurate eye detection, then recognition accuracy improves, but the system becomes more sensitive to image quality requirements
Solution Approach 1:
The patent transforms the detection approach from spatial frequency analysis (sensitive to image quality) to temporal frequency analysis of color components. By analyzing the spectral distribution of red, green, and blue color statistics across time frames, the system achieves adaptability to various image qualities while maintaining detection accuracy
Solution Approach 2:
The patent adds a temporal dimension to the analysis by processing color statistic information across multiple time frames. This time-domain transformation creates an additional dimension for feature extraction, enabling the system to detect human characteristics that persist over time regardless of spatial image quality
Data Source
AI summary
A human face recognition method and apparatus are provided. A processor of the human face recognition apparatus calculates red, green, and blue component statistic information for each of a plurality of human face images. The processor uses an independent component analysis algorithm to analyze component statistic information of two colors and derive a piece of first component information and a piece of second component information. The processor transforms the pieces of first component information and second component information into a frequency domain to derive a piece of first frequency-domain information and a piece of second frequency-domain information. The processor calculates an energy value of the first frequency-domain information within a frequency range. The energy value is used to decide whether the human face images are captured from a human being.