KEYWORDS: Imaging arrays, Image processing, 3D displays, 3D surface sensing, Data processing, Object recognition, Integral imaging, 3D image processing, Photodetectors, Sensing systems
Integral imaging (II) combined with photon-counting detection has been researched for three-dimensional (3D) information sensing under low-light-level condition. This paper addresses the nonlinear correlation of photon-counting integral imaging. The first and second order statistical properties of the nonlinear correlation are verified with varying the mean number of photo-counts in the scene. In the experiment, various nonlinearity factors are tested and simulation results are compared with the theoretically driven results.
Infrared (IR) imaging has the capability to detect thermal characteristics of objects under low-light conditions. This paper addresses IR image segmentation with Gaussian mixture modeling. An IR image is segmented with Expectation Maximization (EM) method assuming the image histogram follows the Gaussian mixture distribution. Multi-level segmentation is applied to extract the region of interest (ROI). Each level of the multi-level segmentation is composed of the k-means clustering, the EM algorithm, and a decision process. The foreground objects are individually segmented from the ROI windows. In the experiments, various methods are applied to the IR image capturing several humans at night.
In this paper, we review automatic concealed object recognition with multi-channel passive millimeter wave images. A four-channel passive millimeter wave imaging system operates in the 8 and 3 mm wavelength regimes with linear vertical and horizontal polarization directions. Registration between multi-channel images and segmentation of concealed objects are addressed. Multi-channel image registration is performed by means of the affine transform derived by the geometric feature matching. Gaussian mixture models are adopted to cluster hidden object pixels in the images. Multi-level segmentation separates the human body region from the background, and concealed objects from the body region, sequentially. In the experiments, the metallic and non-metallic objects concealed under clothing are captured and processed.
KEYWORDS: Image fusion, Cameras, Facial recognition systems, Principal component analysis, Data fusion, Surveillance, Sensors, Silicon, Point spread functions, Communication engineering
Face classification of multiple cameras has wide applications in surveillance. In this paper, the efficacy of a multi-frame
decision-level fusion scheme for face classification based on the photon-counting linear discriminant analysis is
investigated. The photon-counting linear discriminant analysis method is able to realize Fisher’s criterion without
preprocessing for dimensionality reduction. The decision-level fusion scheme is comprised of three stages: score
normalization, score validation, and score combination. After normalization, the candidate scores are selected and
combined by means of a score validation process and a fusion rule, respectively, in order to generate a final score. In the
experiments, out-of-focus and motion blurs are rendered on test images simulating harsh conditions.
Face detection at a distance is very challenging because the image quality becomes low. This paper discusses a face
detection method in the long distance with AdaBoost filtering and a false alarm reduction scheme. The false alarm
reduction scheme is based on skin-color testing and variable edge mask filtering. The skin-color test involves the average
RGB components of the window, followed by the binary cluster image generation. The binary cluster is composed of the
alternative and null pixels according to color. The size of the edge mask is determined by the ellipse covering the binary
cluster. The edge mask filters out false alarms by evaluating the contour shape of the object in the window. In the
experiments, the false alarm reduction scheme is shown to be effective for face detection in images captured at a distance.
Millimeter waves imaging draws increasing attention in security applications for weapon detection under clothing. In this
paper, concealed object segmentation and three-dimensional localization schemes are reviewed. A concealed object is
segmented by the k-means algorithm. A feature-based stereo-matching method estimates the longitudinal distance of the concealed object. The distance is estimated by the discrepancy between the corresponding centers of the segmented
objects. Experimental results are provided with the analysis of the depth resolution.
Passive millimeter wave imaging is useful for security applications since it can detect objects concealed under clothing. However, because of the diffraction limit and low signal level, the automatic image analysis is very challenging. The multi-level segmentation of passive millimeter wave images is discussed as a way to detect concealed objects under clothing. Our passive millimeter wave imaging system is equipped with a Cassegrain dish antenna and a receiver channel operating around 3 mm wavelength. The expectation-maximization algorithm is adopted to cluster pixels on the basis of a Gaussian mixture model. The multi-level segmentation is investigated with more than two clusters to recognize the hidden object in different parts. The performance is evaluated by the average probability error. Experiments confirm that the presented method is able to detect the wood grip of a hand ax as well as the metal part concealed under clothing.
Infrared imaging allows surveillance during the night, thus it has been widely used for military and security applications.
However, infrared images are generally characterized by low resolution, low contrast, and an unclear texture with no
color information. Moreover, various types of noises and background clutters can degrade the image quality. This paper
discusses multi-level segmentation for infrared images. The expectation-maximization algorithm is adopted to cluster
pixels on the basis of Gaussian mixture models. The use of the multi-level segmentation method enables the extraction of
human target regions from the background of the image. Several infrared images are processed to demonstrate the
effectiveness of the presented method.
Millimeter wave (MMW) readily penetrates fabrics, thus it can be used to detect objects concealed under clothing. A passive MMW imaging system can operate as a stand-off type sensor that scans people both indoors and outdoors. However, because of the diffraction limit and low signal level, the imaging system often suffers from low image quality. Therefore, suitable computational processing would be required for automatic analysis of the images. The authors present statistical and computational algorithms and their implementations for real-time concealed object detection. The histogram of the image is modeled as a Gaussian mixture distribution, and hidden object areas are segmented by a multilevel scheme involving the expectation-maximization algorithm. The complete algorithm has been implemented in both MATLAB and C++. Experimental and simulation results confirm that the implemented system can achieve real-time detection of concealed objects.
Millimeter wave (MMW) imaging is finding rapid adoption in security applications such as concealed object detection
under clothing. A passive MMW imaging system can operate as a stand-off type sensor that scans people in both indoors
and outdoors. However, the imaging system often suffers from the diffraction limit and the low signal level. This paper
discusses real-time concealed object recognition based on geometric descriptors. The concealed object region is
extracted by the multi-level segmentation method. A novel approach is proposed to measure similarity between a true
object model and segmented binary objects. Principal component analysis (PCA) regularizes the shape in terms of
translation and rotation. Size normalization provides scale-invariant property. A geometric feature vector is composed of
several shape descriptors. The feature vector is invariant to scale, rotation, and translation, and tolerant to distortion.
Classification is performed by means of measuring Euclidean distance between the mean feature vector of training
models and the feature vector of the segmented object. Experiments confirm that the proposed method provides fast and
reliable recognition of the concealed object carried by a moving human subject.
KEYWORDS: Facial recognition systems, Linear filtering, Phase only filters, Image filtering, Mobile devices, Optical filters, System identification, Fourier transforms, Human subjects, Java
Face identification at a distance is very challenging since captured images are often degraded by blur and noise.
Furthermore, the computational resources and memory are often limited in the mobile environments. Thus, it is very
challenging to develop a real-time face identification system on the mobile device. This paper discusses face
identification based on frequency domain matched filtering in the mobile environments. Face identification is performed
by the linear or phase-only matched filter and sequential verification stages. The candidate window regions are decided
by the major peaks of the linear or phase-only matched filtering outputs. The sequential stages comprise a skin-color test
and an edge mask filtering test, which verify color and shape information of the candidate regions in order to remove
false alarms. All algorithms are built on the mobile device using Android platform. The preliminary results show that
face identification of East Asian people can be performed successfully in the mobile environments.
Passive millimeter wave imaging is very useful for security applications since it candetect objects concealed under
clothing. In this paper,the multi-level segmentation of passive millimeter wave images is presented to detectconcealed
objects under clothing. Our passive millimeter wave imaging system is equipped with a Cassegrain dish antenna and a
receiver channel operating around 3 mm wavelength. The expectation-maximization algorithm is adopted to cluster
pixelson the basis ofa Gaussian mixture model. The multi-level segmentation is investigated with different numbers of
clusters in Gaussian mixture distribution. The performance is evaluated by average probability error.
Experimentsconfirm that the presented method is able to detect the wood grip as well as metal part of the hand
axconcealed under clothing.
The distortions in the perceived images of three different camera arrangements of parallel, converging and diverging are
compared for different inter-camera distances. The parallel also reveals distortion in both depth and image size when
inter-camera distance is small. The converging and diverging are showing opposite behaviors in their distortion
appearance. The distortions are most for the diverging but it can enhance the protruding depth, especially when the intercamera
distance is much smaller than the interocular distance.
Face classification in an uncontrolled setting has wide applications in security and surveillance systems. Multiple frames
are often available for this purpose captured by multiple sensors or a single sensor generating video clips. Data fusion
technique for face classification has an advantage in that a considerable amount of information can be used to achieve
high recognition performance. This paper investigates the efficacy of multi-frame decision level fusion for face
classification based on a photon-counting linear discriminant analysis, which realizes Fisher's criterion without
dimensionality reduction. Decision level fusion comprises two stages: score validation and score combination. During
score validation, candidate symbols (classes) are selected by a screening process. During score combination, the
candidate scores are combined in order to make a final decision. In the experiments, a facial image database is employed
to show the preliminary results of the proposed technique.
A diverging-type stereo camera arrangement is introduced to use in hand-held mobile devices such as mobile phone,
hand PC and introscopes. The arrangement allows making the
inter-camera distance much smaller than that in the
conventional stereo camera arrangements such as parallel and radial types by adjusting the diverging angle. Computer
simulation shows that it can introduce more distortion than the parallel type but it can enhance the depth sense.
We address three-dimensional passive millimeter-wave imaging (MMW) and depth estimation for remote objects. The
MMW imaging is very useful for the harsh environment such as fog, smoke, snow, sandstorm, and drizzle. Its
penetrating property into clothing provides a great advantage to security and defense systems. In this paper, the featurebased
passive MMW stereo-matching process is proposed to estimate the distance of the concealed object under clothing.
It will be shown that the proposed method can estimate the distance of the concealed object.
We address an image registration and segmentation method to detect concealed objects captured by passive millimeter
wave (MMW) imaging. Passive MMW imaging can create interpretable imagery on the objects concealed under clothing.
Due to the penetrating property of the MMW imaging, the MMW imaging system is often employed for the security and
defense system. In this paper, we utilize a multi-channel PMMW imaging system operating at the 8 mm regime with
linear polarization. Image registration and segmentation are performed to detect concealed objects under clothing. The
registration is preceded to align different channel images by means of geometric feature extraction and a matching
process. The Linde-Buzo-Gray (LBG) vector quantization with multi-channel information is adopted to segment the
concealed object from the body area. In the experiment, the automated image registration and segmentation are
performed with various concealed objects including a metal axe and a liquid container.
This paper addresses the registration and the fusion techniques between passive millimeter wave (MMW) and visual
images for concealed object detection. The passive MMW imaging system detects concealed objects such as metal and
man-made objects as well as small liquid and gel containers. The registration and fusion processes are required to
combine information from both of visual and MMW images. The registration process is composed of feature extraction
and matching stages. The body areas in two images are adjusted to each other in scale, rotation, and location. The image
fusion method is based on discrete wavelet transform and a fusion rule, which emphasizes the person's identity and the
hidden object together. The experimental and simulation results show the proposed technique can detect a concealed
object and fuse two different types of images in a fully automated way.
A focal plane detector array in a millimeter wave imaging system can be used to acquire multiview
images in millimeter wave band. Two focal plane detectors which are distanced 8mm are used to obtain a
stereoscopic image pair of a scene. The pair reveals a good depth sense though its resolution is very low
and enables to estimate distances of objects in the scene with a reasonable accuracy.
Keywords: millimeter wave imaging system, parabolic antenna, stereoscopic image pair, focal plane
detector array, depth sense, object distance.
KEYWORDS: Nonlinear filtering, Photon counting, Image filtering, 3D image processing, 3D image reconstruction, Digital filtering, 3D acquisition, Image processing, Automatic target recognition, Linear filtering
In this paper we overview the nonlinear matched filtering for photon counting recognition with 3D passive sensing. The
first and second order statistical properties of the nonlinear matched filtering can improve the recognition performance
compared to the linear matched filtering. Automatic target reconstruction and recognition are addressed for partially
occluded objects. The recognition performance is shown to be improved significantly in the reconstruction space. The
discrimination capability is analyzed in terms of Fisher ratio (FR) and receiver operating characteristic (ROC) curves.
In this paper, we address distortion-tolerant object recognition using photon-counting three-dimensional (3D) integral
imaging (II). A photon-counting linear discriminant analysis (LDA) is reviewed for classification of out-of-plane rotated
objects. We also investigate the effect of a large number of photons and the irradiance change in training and test objects.
In this keynote address, we address three-dimensional (3D) distortion-tolerant object recognition using photon-counting
integral imaging (II). A photon-counting linear discriminant analysis (LDA) is discussed for classification of photon-limited
images. We develop a compact distortion-tolerant recognition system based on the multiple-perspective imaging
of II. Experimental and simulation results have shown that a low level of photons is sufficient to classify out-of-plane
rotated objects.
KEYWORDS: Microorganisms, Digital holography, Image segmentation, 3D image processing, Holograms, 3D image reconstruction, Holography, Microscopy, Statistical analysis, 3D visualizations
We address the optical system for three-dimensional (3D) sensing, visualization and recognition of biological microorganisms. A digital holographic microscopy records Fresnel digital hologram of the biological microorganism. 3D image of the biological microorganism is computationally reconstructed by inverse Fresnel transformation of the digital hologram. For 3D recognition, two methods are presented. One is 3D morphology-based recognition and the other is based on statistical estimation and inference algorithms.
In this keynote address, we introduce three-dimensional (3D) sensing, visualization and recognition of microorganisms using microscopy-based single-exposure on-line (SEOL) digital holography. A coherent Mach-Zehnder interferometer records Fresnel diffraction field by a single on-line exposure to generate a microscopic digital hologram. Complex amplitude distribution is numerically reconstructed by the inverse Fresnel transform at arbitrary depth planes. After the reconstruction of volumetric complex images, 3D biological micro-objects are segmented and features are extracted by Gabor-based wavelets. The graph matching technique searches predefined 3D morphological shapes of reference biological microorganisms. Preliminary experimental results using sphacelaria alga and tribonema aequale alga are presented.
In this keynote address, we introduce three-dimensional (3D) passive sensing using photon counting integral imaging. We investigate both linear and nonlinear matched filtering for automatic target recognition (ATR). Significant benefits of the nonlinear matched filtering with 3D integral imaging are found for ATR with a low number of photons. The discrimination capability of our system is quantified in terms of discrimination ratio (DR), Fisher ratio (FR), and receiver operating characteristic (ROC) curves. Experimental and simulation results are presented.
We propose automated identification of microorganisms using three-dimensional (3-D) complex morphology. This 3-D complex morphology pattern includes the complex amplitude (magnitude and phase) of computationally reconstructed holographic images at arbitrary depths. Microscope-based single-exposure on-line (SEOL) digital holography records and reconstructs holographic images of the biological microorganisms. The 3-D automatic recognition is processed by segmentation, feature extraction by Gabor-based wavelets, automatic feature vector selection by graph matching, training rules, and a decision process. Graph matching combined with Gabor feature vectors measures the similarity of complex geometrical shapes between a reference microorganism and unknown biological samples. Automatic selection of the training data is proposed to achieve a fully automatic recognition system. Preliminary experimental results are presented for 3-D image recognition of Sphacelaria alga and Tribonema aequale alga.
KEYWORDS: 3D image processing, Image compression, 3D displays, Signal to noise ratio, Image quality, LCDs, Optical engineering, Integral imaging, 3D image reconstruction, Image resolution
We present a method to compress ray information in three-dimensional (3-D) integral imaging (II) using the Karhunen-Loeve transform (KLT). Elemental images in II are highly correlated because they are picked-up by numerous lenslets, thus, the KLT can compress an integral image more effectively than 2-D ordinary images. In the hybrid coding scheme, the KLT is distinctly applied to disjoint subsets which are partitioned by a vector quantization (VQ). In the optical experiments, we show that 3-D image quality degradation is negligible if approximately 10% of eigenvectors with the largest eigenvalues are used. We also evaluate the quality of uncompressed images by signal-to-noise ratio (SNR) and peak-to-peak signal-to-noise ratio (PSNR). The presented coding scheme is shown to be better than JPEG (Joint Photographic Experts Group) for the lower bit rates in the II compression. Optical and numerical experiments are presented.
KEYWORDS: Object recognition, 3D image processing, Simulation of CCA and DLA aggregates, Optical engineering, X-ray imaging, Wavelets, 3D modeling, X-rays, Feature extraction, Distortion
This paper presents a distortion-tolerant 3-D volume object recognition technique. Volumetric information on 3-D objects is reconstructed by x-ray imaging. We introduce 3-D feature extraction, volume matching, and statistical significance testing for the 3-D object recognition. The 3D Gabor-based wavelets extract salient features from 3-D volume objects and represent them in the 3-D spatial-frequency domain. Gabor coefficients constitute feature vectors that are invariant to translation, rotation, and distortion. Distortion-tolerant volume matching is performed by a modified 3-D dynamic link association (DLA). The DLA is composed of two stages: rigid motion of a 3-D graph, and elastic deformation of the graph. Our 3-D DLA presents a simple and straightforward solution for a 3-D volume matching task. Finally, significance testing decides the class of input objects in a statistical manner. Experiment and simulation results are presented for five classes of volume objects. We test three classes of synthetic data (pyramid, hemisphere, and cone) and two classes of experimental data (short screw and long screw). The recognition performance is analyzed in terms of the mean absolute error between references and input volume objects. We also confirm the robustness of the recognition algorithm by varying system parameters.
Object recognition and identification is one of essential parts for Homeland Security. There have been numerous researches dealing with object recognition using two-dimensional (2D) or three-dimensional (3D) imaging. In this paper, we address 3D object classification with computational holographic imaging. A 3D object can be reconstructed at different planes using a single hologram. We apply Principal Component Analysis (PCA) and Fisher Linear Discriminant (FLD) analysis based on Gabor-wavelet feature vectors to classify 3D objects measured by digital interferometry. The presented technique substantially reduces the dimensionality of the 3D classification problem. Experimental and simulation results are presented for regional filtering concentrated at specific positions, and for overall grid filtering.
This paper deals with 3D object classification using computational holographic imaging. A 3D object can be reconstructed at different planes using a single hologram. We apply Principal Component Analysis (PCA) and Fisher Linear Discriminant (FLD) analysis based on Gabor-wavelet feature vectors to classify 3D objects measured by digital interferometry. Experimental and simulation results are presented for regional filtering concentrated at specific positions, and for overall grid filtering. The proposed technique substantially reduces the dimensionality of the 3D classification problem.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.