Poincare Index Method for Fingerprint Core Point Extraction

The Poincare Index method is a classical approach for fingerprint core point extraction, primarily used to locate core points in fingerprint images. Its core concept relies on the topological characteristics of the fingerprint orientation field, identifying singular points (such as core points and delta points) by calculating the total directional change along closed paths in the orientation field.

The method operates through three key steps: Orientation Field Calculation: First, preprocessing is applied to the fingerprint image to estimate the local ridge orientation for each pixel, forming the orientation field. Common techniques include gradient-based methods (using horizontal and vertical derivatives with Sobel or Prewitt operators) or frequency-domain analysis (employing Fourier transform for spectral analysis). Poincare Index Calculation: A closed path (e.g., small square or circle) is selected in the orientation field, and the cumulative directional angle changes along adjacent points on the path are computed. When the total angular change equals ±180°, the closed path may contain a core point (+180°) or a delta point (−180°). Implementation typically involves iterating through path segments using modulo arithmetic to handle angle circularity. Core Point Verification: Multi-scale closed paths or orientation field consistency checks are applied to eliminate noise interference, ultimately determining the precise coordinates of core points. This may involve algorithms like region growing or confidence-based filtering.

The method demonstrates strong robustness for low-quality fingerprint images but suffers from high computational complexity. Practical implementations often incorporate orientation field smoothing (Gaussian filtering) or machine learning optimization strategies (e.g., SVM classification) to improve efficiency.

Extension Insight: The Poincare Index method can also be adapted for singular point detection in other biometric features (e.g., palmprints), serving as a fundamental framework for understanding topological feature extraction in pattern recognition systems.