This thesis describes a face recognition system that overcomes the problem of changes in gesture and mimics in three-dimensional (3D) range images. Here, we propose a local variation detection and restoration method based on the two-dimensional (2D) principal component analysis (PCA). The depth map of a 3D facial image is first smoothed using median filter to minimize the local variation. The detected face shape is cropped & normalized to a standard image size of 101x101 pixels and the forefront nose point is selected to be the image center. Facial depthvalues are scaled between 0 and 255 for translation and scaling-invariant identification. The preprocessed face image is smoothed to minimize the local variations. The 2DPCA is applied to the resultant range data and the corresponding principal-(or eigen-) images are used as the characteristic feature vectors of the subject to find his/her identity in the database of pre-recorded faces.
The system's performance is tested against the GavabDB facial databases. Experimental results show that the proposed method is able to identify subjects with different gesture and mimics in the presence of noise in their 3D facial image.
Table of Contents
I. INTRODUCTION
II. BACK GROUND AND REVIEW OF PAST WORK
III. MAIN WORK
IV. IMPLEMENTATION & RESULTS
A. Data base
B. Experiment results and matching
V. CONCLUSION
Research Objectives and Themes
The primary objective of this research is to develop a robust face recognition system for 3D range images that effectively addresses the challenges posed by variations in facial expressions and gestures, thereby minimizing recognition errors through advanced image preprocessing and principal component analysis.
- Development of a 3D facial image preprocessing pipeline involving noise reduction and normalization.
- Application of Two-Dimensional Principal Component Analysis (2DPCA) for efficient feature extraction.
- Implementation of Euclidean distance-based classification for identity verification.
- Evaluation of system performance using the GavaDB facial database under various conditions.
Excerpt from the Publication
III. MAIN WORK
In this research effort, we propose a new modeling technique for human face recognition problem in 3D images with the ability of encompassing different expressions of one's facial image.The proposed method is illustrated by a block diagram representation in Figure 1. A facial image is first subjected to a preprocessing stage in which, the surface smoothing stage is carried out to prevent variations giving false matching results. The resultant facial image map is normalized and then aligned with the coordinate system centered at the detected nose point. Then the background is eliminated by croping the depth map of the 3D image. Then Each image is described by its principal components using 2DPCA. The resultant principal components are used as feature vectors in the classification stage to calculate the similarity between two facial images. The Euclidean distance classifier is used in the matching process. Range images are a special class of digital images. Each pixel of a range image expresses the distance between a known reference frame and a visible point in the scene. Therefore, a range image reproduces the 3D structure of a scene.
Range images are also referred to as depth images, depth maps, xyz maps, surface profiles and 2.5D images.
Range images can be represented in two basic forms. One is a list of 3D coordinates in a given reference frame (cloud of points), for which no specific order is required. The other is a matrix of depth values of points along the directions of the x,y image axes, which makes spatial organization explicit.
Summary of Chapters
I. INTRODUCTION: Outlines the challenges of facial recognition in 3D, specifically focusing on the impact of facial expressions and gestures on recognition accuracy.
II. BACK GROUND AND REVIEW OF PAST WORK: Provides a literature review of existing 2D and 3D face recognition methodologies and identifies limitations in current approaches regarding illumination and pose invariance.
III. MAIN WORK: Details the proposed algorithmic framework, including the preprocessing steps and the application of 2DPCA for feature extraction and classification.
IV. IMPLEMENTATION & RESULTS: Describes the experimental setup using the GavaDB database and provides a performance analysis of the system based on different median filter window sizes.
V. CONCLUSION: Summarizes the findings and effectiveness of the proposed 3D face recognition method and suggests directions for future improvements regarding pose correction.
Keywords
3D Face Recognition, Principal Component Analysis, 2DPCA, Range Images, Median Filter, Facial Expression, Euclidean Distance, Eigen Faces, Image Preprocessing, GavaDB, Feature Extraction, Depth Map, 3D Mesh, Pattern Recognition, Biometrics
Frequently Asked Questions
What is the core focus of this research?
The research focuses on developing an efficient face recognition system capable of handling 3D range images, specifically addressing errors caused by facial expressions and varying gestures.
What are the central themes of the study?
The study centers on 3D image preprocessing, surface smoothing, depth map normalization, and the application of principal component analysis to identify unique facial features.
What is the primary objective of this work?
The primary objective is to improve the reliability of 3D face recognition systems by using 2DPCA to extract feature vectors that remain robust against internal facial deformations.
Which scientific methods are utilized?
The research employs median filtering for noise reduction, image cropping/alignment techniques, 2DPCA for dimension reduction, and Euclidean distance classifiers for matching.
What does the main body of the text cover?
The main body covers the theoretical framework, the algorithmic stages of the proposed model, implementation details, and a quantitative analysis of recognition rates.
Which keywords characterize this publication?
Key terms include 3D Face Recognition, 2DPCA, Range Images, Eigen Faces, and Facial Expression analysis.
How does the proposed system handle noise in 3D images?
The system applies a median filter, specifically of size 5x5, to smooth the depth map and eliminate spikes or artifacts around the nose area.
What role does the GavaDB database play in this research?
The GavaDB database is used as the benchmark dataset containing 427 3D facial images to evaluate the performance and recognition accuracy of the proposed algorithm.
What is the significance of the 101x101 pixel normalization?
Normalization to a standard 101x101 pixel size ensures that all facial images are consistent in scale, allowing for accurate comparison during the feature extraction stage.
What are the limitations of the proposed approach?
The authors acknowledge that the current system lacks a formal pose correction stage and is incapable of treating major artifacts resulting from poor sensing conditions.
- Arbeit zitieren
- Yagnesh Parmar (Autor:in), 2012, Efficient 3D face recognition based on PCA, München, GRIN Verlag, https://www.hausarbeiten.de/document/203464
