Two-dimensional face-recognition techniques suffer from facial texture and illumination variations. Although 3-D techniques can overcome these limitations, the reconstruction and storage expenses of 3-D information are extremely high. We present a novel face-recognition method that directly utilizes 3-D information encoded in face fringe patterns without having to reconstruct 3-D geometry. In the proposed method, a digital video projector is employed to sequentially project three phase-shifted sinusoidal fringe patterns onto the subject's face. Meanwhile, a camera is used to capture the distorted fringe patterns from an offset angle. Afterward, the face fringe images are analyzed by the phase-shifting method and the Fourier transform method to obtain a spectral representation of the 3-D face. Finally, the eigenface algorithm is applied to the face-spectrum images to perform face recognition. Simulation and experimental results demonstrate that the proposed method achieved satisfactory recognition rates with reduced computational complexity and storage expenses.
In this paper, a quality-guided phase unwrapping method is proposed for a modified Fourier transform method, which
utilizes a fringe image and a flat image. The proposed method takes advantage of the additional information provided by
the flat image to calculate the visibility of the fringe pattern, which is difficult to obtain in the conventional Fourier
transform method. The phase unwrapping process includes two steps. First, the pixels with unreliable phase values in the
wrapped phase map are masked out. Then, visibility is chosen as a quality index to assist the quality-guided unwrapping.
Experimental results show that the proposed method is an effective method to unwrap complex surface's phase map
generated from dense fringe patterns.
The phase-to-coordinate conversion is an important step for accurate 3-D shape measurement by use of structured light methods. We propose to use an absolute phase map in the Fourier transform method to achieve better accuracy in the phase-to-coordinate conversion. A cross-shaped marker is embedded in the fringe pattern that is projected onto the object. The position of the marker in the captured fringe image is detected and utilized to calculate the absolute phase map. For phase analysis of the fringe image, the marker is removed and the sinusoidal intensity distribution of the fringe pattern is restored before the forward and inverse Fourier transforms are applied. Experimental results of absolute phase retrieval and 3-D reconstruction are presented to show the feasibility of the proposed method.
The difficulty of implementing the phase shifting method in shadow moiré lies in the fact that the phase shift due to the displacement of the light source, the imaging sensor, or the grating is non-uniform across the field of view. Typical phase shifting algorithms fail to produce accurate results. In the past few decades, various approximation methods have been developed to overcome this difficulty. In this paper, we describe an elegant solution that provides exact close-form result. In our proposed system, the grating is translated in equal steps to introduce phase shifts. The phase value at each point is determined by the Carré algorithm, which only requires uniform phase shifts for each point, instead of in the whole field of view. The 3-D shape of the object is then reconstructed from the phase map retrieved from the Carré algorithm. The simulation results demonstrate the effectiveness of the Carré algorithm for shadow moiré.
In this paper, a novel modified Fourier transform method is proposed, which employs a fringe image and a flat image to eliminate the background and in the mean time facilitate the retrieval of the absolute phase map. Both the fringe and flat patterns are projected onto the object by a digital video projector. With the subtraction of the flat image from the fringe image, the background is completely removed and the spectrum overlapping in the frequency domain is prevented. The flat image is also employed for hole and shadow detection. Two cross-shaped markers are embedded in the flat and fringe image respectively for absolute phase retrieval. Experimental results showed that the proposed method produced better shape measurement results when measuring fast moving or changing objects, compared to the phase shifting method. The proposed method has the potential to boost the speed of our real-time 3-D shape measurement system to 120 fps with better measurement accuracy.
Accurate 3D shape measurement via fringe analysis methods requires the determination of the absolute phase map of the
object. In this paper, we present a novel method for absolute phase retrieval developed for use with the Fourier transform
method for fringe analysis. A cross-shaped marker is embedded in the fringe pattern that is projected to the object. The
position of the marker in the captured fringe image is detected and later used in calculating the absolute phase map. For
phase analysis of the fringe image, the marker is removed and the sinusoidal intensity distribution of the fringe pattern is
restored before the Fourier transform method is applied. This paper focuses on the concept of absolute phase retrieval
from a single fringe pattern as well as techniques on marker detection and removal. Experimental results on absolute
phase retrieval and 3D reconstruction are also presented to show the feasibility of the proposed method.