Face Feature Extraction Algorithm Using Non-negative Matrix Factorization (NMF)

MATLAB 1K 226 views 0 downloads 1 credits
Tags:

Resource Overview

Face feature extraction algorithm based on Non-negative Matrix Factorization (NMF), where the fundamental concept involves finding a linear subspace W such that all pixel values in the basis images constituting this subspace are positive, and the projection coefficients of face images onto this subspace are also non-negative.

Detailed Documentation

The face feature extraction algorithm based on Non-negative Matrix Factorization (NMF) works by identifying a linear subspace W composed of basis images with strictly positive pixel values. Face images projected onto this subspace yield non-negative coefficients, enabling effective feature extraction for applications like face recognition and facial expression analysis.

Implementation typically involves optimizing the objective function using multiplicative update rules or gradient descent methods, where the input face image matrix V is factorized into two non-negative matrices W (basis images) and H (coefficient matrix) such that V ≈ WH. Key functions in MATLAB implementations include norm() for error calculation and iterative updates using element-wise operations to maintain non-negativity constraints.

You May Also Like