Robust Head Pose Estimation Using Contourlet Transform

TL;DR

This paper introduces a robust head pose estimation method utilizing contourlet transform for feature extraction, combined with dimensionality reduction and classification techniques, achieving superior accuracy on the FERET database.

Contribution

The novel approach integrates contourlet transform with LDA/PCA and classifiers for improved head pose estimation accuracy.

Findings

Demonstrates robustness on FERET database

Outperforms existing methods in accuracy

Effective feature reduction and classification pipeline

Abstract

Estimating pose of the head is an important preprocessing step in many pattern recognition and computer vision systems such as face recognition. Since the performance of the face recognition systems is greatly affected by the poses of the face, how to estimate the accurate pose of the face in human face image is still a challenging problem. In this paper, we represent a novel method for head pose estimation. To enhance the efficiency of the estimation we use contourlet transform for feature extraction. Contourlet transform is multi-resolution, multi-direction transform. In order to reduce the feature space dimension and obtain appropriate features we use LDA (Linear Discriminant Analysis) and PCA (Principal Component Analysis) to remove ineffcient features. Then, we apply different classifiers such as k-nearest neighborhood (knn) and minimum distance. We use the public available FERET database to evaluate the performance of proposed method. Simulation results indicate the superior robustness of the proposed method.