This PDF file contains the front matter associated with SPIE Proceedings Volume 7877, including the Title Page, Copyright information, Table of Contents, Introduction, and the Conference Committee listing.

In this work, we attempt to propose a signal restoration technique from the noise corrupted signal. The main diculty in most of the noise removal approaches is the extraction of singularities which are part of the signal from noise elements. In order to over come this problem, the propose method measures the Lipschitz exponent of the transitions to extract the noise elements. Unlike many noise removal techniques, the present method works in the non orthogonal domain. These noise elements were identied from the decaying slope of modulus maxima lines and is termed as Lipschitz exponents. The main contribution of the work is the reconstruction process. By utilizing the property of Lipschitz exponents, it is possible to reconstruct the smooth signal by non linear functioning. Statistical results are quite promising and performs better than conventional shrinkage methods in the case of high variance noise. Furthermore, in order to extract noise elements the proposed method is not limited with the selection of wavelet function for the addressed signal as well.

Automatic banknote sheet cut-and-bundle machines are widely used within the scope of banknote production. Beside the cutting-and-bundling, which is a mature technology, image-processing-based quality inspection for this type of machine is attractive. We present in this work a new real-time Touchless Counting and perspective cutting blade quality insurance system, based on a Color-CCD-Camera and a dual-core Computer, for cut-and-bundle applications in banknote production. The system, which applies Wavelet-based multi-scale filtering is able to count banknotes inside a 100-bundle within 200-300 ms depending on the window size.

This paper is focused on the characterization of GNSS signals reception environment by estimation of the percentage of visible sky in real-time. On previous works, a new segmentation technique based on a color watershed using an adaptive combination of color and texture information was proposed. This information was represented by two morphological gradients, a classical color gradient and a morphological texture gradient based on mathematical morphology or co-occurrence matrices. The segmented images were then classified into two regions: sky and not-sky. However, this approach has high computational cost and thus, cannot be applied in real-time. On this paper, we present this adaptive segmentation method with a texture gradient calculated by the Gabor filter and a region-tracking method based on a block-matching estimation. This last step reduces the execution time of the application in order to respect the real-time conditions. Since the application works for fish-eye images, a calibration and rectification method is required before tracking and is also presented on this paper. The calibration method presented is based on the straight line condition and thus does not use real word coordinates. This prevents measurement errors. The tracking results are compared to the results of the classification method (which has already been evaluated on previous works). The evaluation shows that the proposed method has a very low error and decreases the execution time by ten times.

The problem of circular object detection and localisation arises quite often in machine vision applications, for example in semi-conductor component inspection. We propose two novel approaches for the precise centre localisation of circular objects, e.g. p-electrodes of light-emitting diodes. The first approach is based on image gradients, for which we provide an objective function that is solely based on dot products and can be maximised by gradient ascend. The second approach is inspired by the concept of isophotes, for which we derive an objective function that is based on the definition of radial symmetry. We evaluate our algorithms on synthetic images with several kinds of noise and on images of semiconductor components and we show that they perform better and are faster than state of the art approaches such as the Hough transform. The radial symmetry approach proved to be the most robust one, especially for low contrast images and strong noise with a mean error of 0.86 pixel for synthetic images and 0.98 pixel for real world images. The gradient approach yields more accurate results for almost all images (mean error of 4 pixel) compared to the Hough transform (8 pixel). Concerning runtime, the gradient-based approach significantly outperforms the other approaches being 5 times faster than the Hough transform; the radial symmetry approach is 12% faster.

Designing a suitable image representation is one of the most fundamental issues of computer vision. There are three steps in the popular Bag of Words based image representation: feature extraction, coding and pooling. In the final step, current methods make an M x K encoded feature matrix degraded to a K-dimensional vector (histogram), where M is the number of features, and K is the size of the codebook: information is lost dramatically here. In this paper, a novel pooling method, based on 2-D histogram representation, is proposed to retain more information from the encoded image features. This pooling method can be easily incorporated into state-of- the-art computer vision system frameworks. Experiments show that our approach improves current pooling methods, and can achieve satisfactory performance of image classification and image reranking even when using a small codebook and costless linear SVM.

This paper presents a new approach to non-rigid registration. A hierarchical subdivision approach is applied, with local normalized phase correlation for patch registration. The major improvement is achieved by implementing a suitable decimation at each level. The decimation is implemented via a Gram polynomial basis. Both global and local polynomial approximation are considered and compared with the use of a Fourier basis. The issue of Gibbs error in polynomial decimation is examined. It is shown that the Gram basis is superior when applied to signals with strong gradient, i.e., a gradient which generates a significant Gibbs error with a Fourier basis. A bivariate Gram polynomial tensor product approximation is used to implement regularization. It is demonstrated that the new method performs well on both synthetic and real image data. The procedure requires approximately 1.3 sec. to register an image with 800 × 500 pixels.

Morphological and microstructural features visible in microscopy images of nuclear materials can give information about the processing history of a nuclear material. Extraction of these attributes currently requires a subject matter expert in both microscopy and nuclear material production processes, and is a time consuming, and at least partially manual task, often involving multiple software applications. One of the primary goals of computer vision is to find ways to extract and encode domain knowledge associated with imagery so that parts of this process can be automated. In this paper we describe a user-in-the-loop approach to the problem which attempts to both improve the efficiency of domain experts during image quantification as well as capture their domain knowledge over time. This is accomplished through a sophisticated user-monitoring system that accumulates user-computer interactions as users exploit their imagery. We provide a detailed discussion of the interactive feature extraction and segmentation tools we have developed and describe our initial results in exploiting the recorded user-computer interactions to improve user productivity over time.

We present an ecient measure of overlap between two co-linear segments which considerably decreases the overall computational time of a Segment-based motion estimation and reconstruction algorithm already exist in literature. We also discuss the special cases where sparse sampling of the motion space for initialization of the algorithm does not result in a good solution and suggest to use dense sampling instead to overcome the problem. Finally, we demonstrate our work on a real data set.

In general, since smoke appears before flames, smoke detection is particularly important for early fire detection systems. To detect fire-smoke using video camera is a difficult work because main characteristics of a smoke are uncertain, vague, constant patterns of shape and color. Thus, this paper proposes a new fire-smoke detection method, especially forest smoke using analyzing of temporal patterns of smoke and Fuzzy Finite Automata (FFA). To consider the smoke characteristics over time, the temporal patterns of intensity entropy, wavelet energy and motion orientation have been used for generating, multivariate probability density functions (PDFs) are applied Fuzzy Finite Automata (FFA) for smoke verification. The proposed FFA consist of a set of fuzzy states (VH, H, L, VL), and a transition mapping that describes what event can occur at which state and resulting new state. For smoke verification, FFA is most appropriate method in case variables are time-dependent and uncertain. The proposed algorithm is successfully applied to various fire-smoke videos and shows a better detection performance.

This paper presents a new approach for the estimation of fire front volume in indoor laboratory experiments. This work deals with fire spreading on inclinable tables. The method is based on the use of two synchronized stereovision systems positioned respectively in a back position and in a front position of the fire propagation direction. The two vision systems are used in order to extract complementary 3D fire points. The obtained data are projected in a same reference frame and used to build a global form of the fire front. An inter-systems calibration procedure is presented and permits the computation of the projection matrix in order to project all the data to a unique reference frame. From the obtained 3D fire points, a three dimensional surface rendering is performed and the fire volume is estimated.

Automatic recognition of road distresses has been an important research area since it reduces economic loses before cracks and potholes become too severe. Existing systems for automated pavement defect detection commonly require special devices such as lights, lasers, etc, which dramatically increase the cost and limit the system to certain applications. Therefore, in this paper, a low cost automatic pavement distress evaluation approach is proposed. This method can provide real-time pavement distress detection as well as evaluation results based on the color images captured from a camera installed on a survey vehicle. The entire process consists of two main parts: pavement surface extraction followed by pavement distress detection and classification. In the first part, a novel color segmentation method based on a feed forward neural network is applied to separate the road surface from the background. In the second part, a thresholding technique based on probabilistic relaxation is utilized to separate distresses from the road surface. Then, by inputting the geometrical parameters obtained from the detected distresses into a neural network based pavement distress classifier, the defects can be classified into different types. Simulation results are given to show that the proposed method is both effective and reliable on a variety of pavement images.

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.

Unconstrained environments with variable ambient illumination and changes of head pose are still challenging for many face recognition systems. To recognize a person independent of pose, we first fit an active appearance model to a given facial image. Shape information is used to transform the face into a pose-normalized representation. We decompose the transformed face into local regions and extract texture features from these not necessarily rectangular regions using a shape-adapted discrete cosine transform. We show that these features contain sufficient discriminative information to recognize persons across changes in pose. Furthermore, our experimental results show a significant improvement in face recognition performance on faces with pose variations when compared with a block-DCT based feature extraction technique in an access control scenario.

This study proposes a new white blood cell (WBC) segmentation method using region merging scheme and GVF (Gradient Vector Flow) snake. WBC segmentation consists of two schemes; nuclei segmentation and cytoplasm segmentation. For nuclei segmentation, we create a probability map using probability density function estimated from samples of WBC's nuclei and crop the sub-images to include nucleus by using the fact that nuclei have salient color against background and red blood cells. Then, mean-shift clustering is performed for region segmentation and merging rules are applied to merge particle clusters to nucleus. For cytoplasm segmentation, a hybrid approach is proposed that combines the spatial characteristics of cytoplasm and GVF snakes to delineate the boundary of the region of interest. Unlike previous algorithms, the main contribution of this study is to improve the accuracy of WBC segmentation and reduce the computational time by cropping sub-images and applying different segmentation rules according to the parts of cell. The evaluation of proposed method was performed on five WBC types and it showed that the proposed algorithm produced accurate segmentation results in most types of WBCs.

A multi-sensor image fusion method based on redundant lifting non-separable wavelet transform is presented. This method can capture the detailed information of an image accurately. The wavelet coefficients of the approximant sub-images are fused respectively based-on the fusion rule of local area gradient. This method is effectively used to fuse the multi-spectral infrared images and the multi-sensor medical images. The results show that the proposed method has the property of shift-invariance and has good fusion effects.

The detection of abnormalities is a very challenging problem in computer vision, especially if these abnormalities must be detected in images of textured surfaces such as textile, stone, or wood. We propose a novel, non-parametric approach for defect detection in textures that only employs two features. We compute the two parameters of a Weibull fit for the distribution of image gradients in local regions. Then, we perform a simple novelty detection algorithm in order to detect arbitrary deviations of the reference texture. Therefore, we evaluate the Euclidean distances of all local patches to a reference point in the Weibull space, where the reference point is determined for each texture image individually. Thus, our approach becomes independent of the particular texture type and also independent of a certain defect type. For performance evaluation we use the highly challenging database provided by Bosch for a contest on industrial optical inspection with different classes of textures and different defect types. By using the Weibull parameters we can detect local deviations of texture images in an unsupervised manner with high accuracy. Compared to existing approaches such as Gabor filters or grey level statistics, our approach is not only powerful, but also very efficient such that it can also be applied for real-time applications.

At the Spallation Neutron Source (SNS), an accelerator-based neutron source located at the Oak Ridge National Laboratory (Tennessee, USA), the production of neutrons is obtained by accelerating protons against a mercury target. This self-cooling target, however, suffers rapid heat deposition by the beam pulse leading to large pressure changes and thus to cavitations that may be damaging to the container. In order to locally compensate for pressure increases, a small-bubble population is added to the mercury flow using gas bubblers. The geometry of the bubblers being unknown, we are testing several bubblers' configurations and are using machine vision techniques to characterize their efficiency by quantitative measurement of the created bubble population. In this paper we thoroughly detail the experimental setup and the image processing techniques used to quantitatively assess the bubble population. To support this approach we are comparing our preliminary results for different bubblers and operating modes, and discuss potential improvements.

Coded aperture techniques have been applied to neutron radiography to address limitations in neutron flux and resolution of neutron detectors in a system labeled coded source imaging (CSI). By coding the neutron source, a magnified imaging system is designed with small spot size aperture holes (10 and 100μm) for improved resolution beyond the detector limits and with many holes in the aperture (50% open) to account for flux losses due to the small pinhole size. An introduction to neutron radiography and coded aperture imaging is presented. A system design is developed for a CSI system with a development of equations for limitations on the system based on the coded image requirements and the neutron source characteristics of size and divergence. Simulation has been applied to the design using McStas to provide qualitative measures of performance with simulations of pinhole array objects followed by a quantitative measure through simulation of a tilted edge and calculation of the modulation transfer function (MTF) from the line spread function. MTF results for both 100μm and 10μm aperture hole diameters show resolutions matching the hole diameters.

Nowadays machine vision applications require skilled users to configure, tune, and maintain. Because such users are scarce, the robustness and reliability of applications are usually significantly affected. Autonomic computing offers a set of principles such as self-monitoring, self-regulation, and self-repair which can be used to partially overcome those problems. Systems which include self-monitoring observe their internal states, and extract features about them. Systems with self-regulation are capable of regulating their internal parameters to provide the best quality of service depending on the operational conditions and environment. Finally, self-repairing systems are able to detect anomalous working behavior and to provide strategies to deal with such conditions. Machine vision applications are the perfect field to apply autonomic computing techniques. This type of application has strong constraints on reliability and robustness, especially when working in industrial environments, and must provide accurate results even under changing conditions such as luminance, or noise. In order to exploit the autonomic approach of a machine vision application, we believe the architecture of the system must be designed using a set of orthogonal modules. In this paper, we describe how autonomic computing techniques can be applied to machine vision systems, using as an example a real application: 3D reconstruction in harsh industrial environments based on laser range finding. The application is based on modules with different responsibilities at three layers: image acquisition and processing (low level), monitoring (middle level) and supervision (high level). High level modules supervise the execution of low-level modules. Based on the information gathered by mid-level modules, they regulate low-level modules in order to optimize the global quality of service, and tune the module parameters based on operational conditions and on the environment. Regulation actions involve modifying the laser extraction method to adapt to changing conditions in the environment.

We propose a method and a system to measure distances to a target and at the same time obtaining an image of the target. The system is based on a wavefront coded lens and an image processing unit. The method requires a calibration phase where a blur metric is related to the distance of the target. The distance to the target for any position is then obtained by computing the blur metric of the target and using the calibration data. The method and system are easy to manufacture and provide an alternative to other distance measuring methods and devices, while also producing an image of the scene. The target is a printed image of pseudo random black and white elements which can be stick or placed nearby objects of which distance is to be evaluated.

We present a machine vision system for automatic identification of the class of firearms by extracting and analyzing two significant properties from spent cartridge cases, namely the Firing Pin Impression (FPI) and the Firing Pin Aperture Outline (FPAO). Within the framework of the proposed machine vision system, a white light interferometer is employed to image the head of the spent cartridge cases. As a first step of the algorithmic procedure, the Primer Surface Area (PSA) is detected using a circular Hough transform. Once the PSA is detected, a customized statistical region-based parametric active contour model is initialized around the center of the PSA and evolved to segment the FPI. Subsequently, the scaled version of the segmented FPI is used to initialize a customized Mumford-Shah based level set model in order to segment the FPAO. Once the shapes of FPI and FPAO are extracted, a shape-based level set method is used in order to compare these extracted shapes to an annotated dataset of FPIs and FPAOs from varied firearm types. A total of 74 cartridge case images non-uniformly distributed over five different firearms are processed using the aforementioned scheme and the promising nature of the results (95% classification accuracy) demonstrate the efficacy of the proposed approach.

This paper presents a method that can extract and visualize anatomical structures from volumetric medical images by using a 3D level set segmentation method and a hybrid volume rendering technique. First, the segmentation using the level set method was conducted through a surface evolution framework based on the geometric variation principle. This approach addresses the topological changes in the deformable surface by using the geometric integral measures and level set theory. These integral measures contain a robust alignment term, an active region term, and a mean curvature term. By using the level set method with a new hybrid speed function derived from the geometric integral measures, the accurate deformable surface can be extracted from a volumetric medical data set. Second, we employed a hybrid volume rendering approach to visualize the extracted deformable structures. Our method combines indirect and direct volume rendering techniques. Segmented objects within the data set are rendered locally by surface rendering on an object-by-object basis. Globally, all the results of subsequent object rendering are obtained by direct volume rendering (DVR). Then the two rendered results are finally combined in a merging step. This is especially useful when inner structures should be visualized together with semi-transparent outer parts. This merging step is similar to the focus-plus-context approach known from information visualization. Finally, we verified the accuracy and robustness of the proposed segmentation method for various medical volume images. The volume rendering results of segmented 3D objects show that our proposed method can accurately extract and visualize human organs from various multimodality medical volume images.

Many methods have been developed in image processing for face recognition, especially in recent years with the increase of biometric technologies. However, most of these techniques are used on grayscale images acquired in the visible range of the electromagnetic spectrum. The aims of our study are to improve existing tools and to develop new methods for face recognition. The techniques used take advantage of the different spectral ranges, the visible, optical infrared and thermal infrared, by either combining them or analyzing them separately in order to extract the most appropriate information for face recognition. We also verify the consistency of several keypoints extraction techniques in the Near Infrared (NIR) and in the Visible Spectrum.

A multidisciplinary research conducted at the Oak Ridge National Laboratory aims at understanding the molecular controls of partitioning, transport and fate of carbon fixed by photosynthesis in plants and its correlation with other measured plant system properties. Ultimately, we intend to develop a modeling framework to assess, correlate and predict as to which spatiotemporal changes in system dynamics are key to predicting emergent properties of system. Within this research, this paper relates to the quantitative morphological imaging of the main structures forming a plant (stem, roots, and leaves), their internal sub-structures, and changes occurring overtime.

Being a low cost method, the ball crater method is being widely used in both the laboratory and industry for thickness measurements of coatings. A crucial step of this technique is the determination of the radii of the crater circles from a microscope image. The accuracy of the thickness value is mostly defined by the accuracy of radius measurements. Traditional methods of measuring the radii involve human operators, who inevitably introduce measurement uncertainty. In this work, we propose an automated method of estimating the crater radii with the aid of image processing. It measures the radius by fitting a circle to all edge points of the crater circle, which is extracted from the microscope image of the crater by using image processing techniques. It enables automating the process of determining the coating thickness from the image of a ball crater on a coating. Furthermore, because it utilizes all edge points to estimate the radius value, it increases robustness and measurement accuracy. Experimental results confirm the feasibility of our method and its potential in reducing measurement uncertainty and increasing measurement accuracy of the ball crater method.