Search

Xilinx System Generator (XSG) integrates with MATLAB Simulink to build a complex system through Xilinx and Simulink blocks. This paper presents a hardware implementation of various image processing techniques to modify and extract the pixel information from various input images in run time. Complex coding strategies are eradicated with integrated system. The intricacy in structural design is reduced to a huge extent using Xilinx System generator. The reconfigurable attribute of Field Programmable Gate Array (FPGA) bags the massive advantages in image processing like co simulating with selective pixel and verified on FPGA. Image processing algorithms explored in methodical manner by performing diverse operations like edge detection, and color enhancement, complement, contrast stretching etc. The proposed image processing techniques (IPT) intended for various applications. Dissimilar image processing techniques are processed by FPGA and displayed on Video Graphics Array (VGA). Image processing system is designed in XSG, hardware description language (HDL) like VHDL & Verilog code is generated by processing system by system generator token. Design is implemented on Artix-7 and targeted on Basys3 reconfigurable hardware (XC7A35T-1-CPG236C). [ABSTRACT FROM AUTHOR]

The mosaic image is an essential process for merging several images, especially for very high-resolution satellite images, as its width swath is small. The main issue of mosaic images is the geometric accuracy between the images. This paper proposed a sub-pixel geometric evaluation to assess the geometric shift between the images. It is an essential step for image registration. Image registration is a process to determine a geometric shift involving two or more images with the same object but different acquisition dates, viewpoints, and sensors. The basic idea of image registration in the frequency domain is estimating the shift between the test image and the reference image. In this study, we used the phase cross-correlation function to determine the geometric shift between two different acquisition Pleiades images from each polygon. Phase correlation is based on the translation property of the Fourier Transform. It transforms the displacement of two correlated images in the spatial domain into a phase difference in the frequency domain. The phase correlation algorithm calculates a phase difference map containing a single peak. The peak location is proportional to the value of the shift between the two images. Several Pleiades images were used in the experiment. The image processing level is an ortho-image, which is already pan-sharped. As a result, the highest geometric shift is on Polygon 2 based on the R-value, and the lowest error value is in polygon 1. The experiments also showed that the most significant error value is in the blue band, and the lowest is in the red band. The average geometric shift between the two Pleiades data is 1.187 pixels, with the most shift located in the blue band with a 1.282 pixels shift, and the Root Mean Square Error (RMSE) is 1.8. [ABSTRACT FROM AUTHOR]

Mobile cloud technology is currently playing a vital role in the field of mobile communication. Frequently new updates are coming in the field of image processing, online games, Natural language processing and other application using mobile and cloud. On mobile phones, the execution of sophisticated software and Apps might result in poor performance in terms of battery consumption and response time. Therefore a new method is required to utilize the benefits of mobile cloud by using computation offloading technique. In this paper we proposed EDOMAC method for execution of Picasa and Mathdroid application using cloudlet environment. We have computed response time and energy consumption for above mentioned applications based on mobile and cloud. At the end compared results with POMAC and M-POMAC. [ABSTRACT FROM AUTHOR]

Maintaining the integrity of pipelines over long distances is an ongoing challenge for oil and gas companies. Among many factors, corrosion is one of the major causes of pipeline failure. Therefore, timely and accurate detection of corrosion is crucial. Along with the growing use of In-Pipe Inspection Robot (IPIR) technologies, a camera-based visual inspection has become a reliable technique for pipe defects detection. However, the conventional manual surveying process by the operators shows a lack of efficiency in this task. This paper studies the application of YOLO, an image-processing algorithm based on Convolutional Neural Network (CNN), for automating corrosion inspection. The results demonstrate that the proposed method is capable of performing detection with an accuracy rate of 64% under AP75 threshold. The system developed can be a promising tool in providing real-time autonomous defect detection to enhance IPIR devices. [ABSTRACT FROM AUTHOR]

One of the research emergence as per studied problem on the image processing based computer vision is that object detection in a image with bounding boxes. This complicated processing has to be done with help of machine leaning based algorithms only. In the recent years research has done with machine leaning algorithms like CNN,RCNN, Fast RCNN,FasterRCNN,Yolo algorithm and etc.These algorithms have achieved the proposed concept in different levels. In this paper we presented the comparative study of each algorithm and provided efficiency and weakness contexts. [ABSTRACT FROM AUTHOR]

In order to solve the problem of how to reconstruct the original image under the condition of unknown sparse basis, this paper proposes an image reconstruction method based on blind compressed sensing model. In this model, the image signal is regarded as the product of a sparse coefficient matrix and a dictionary matrix. Based on the existing blind compressed sensing theory, the optimal solution is solved by the alternative minimization method. The proposed method solves the problem that the sparse basis in compressed sensing is difficult to represent, which restrains the noise and improves the quality of reconstructed image. This method ensures that the blind compressed sensing theory has a unique solution and can recover the reconstructed original image signal from a complex environment with a stronger self-adaptability. The experimental results show that the image reconstruction algorithm based on blind compressed sensing proposed in this paper can recover high quality image signals under the condition of under-sampling. [ABSTRACT FROM AUTHOR]

In order to solve the ill-posed problem in super-resolution image reconstruction, this paper proposes an adaptive regularization way use sparse representation. We build a new L1/2 non-convex optimization model and apply reweighted L2 Norm for the adaptive algorithm in this paper. Experimental results show the significant effect in denoising and preserving edge details. It outperforms some traditional methods in the value of peak signal to noise ratio and structural similarity [ABSTRACT FROM AUTHOR]

The paper is devoted to solving the problem of excluding areas of irregularities in material surface images. To solve this problem, we propose an algorithm for automatic mask construction with identifying the areas with material, areas without material, and crack areas; this combines image processing methods (pattern search, threshold flood filling, boundary search) with crack detection algorithms. It is demonstrated that this algorithm is applicable to experimental images with a large number of frames in a series. [ABSTRACT FROM AUTHOR]

This paper presents a tool that would calculate traffic density through vehicle detection using image processing. Haar-Cascading Algorithm was used in vehicle detection. The developed tool used video recordings from identified Cross and T-type road junctions in the City of Mandaue. Traffic video recording were provided by the Traffic Enforcement Agency of Mandaue (TEAM). The system will classify traffic status as low, moderate or heavy based on the calculated traffic density. It is also capable of simulating traffic light control by displaying a Go signal in one part of the road in a particular road junction at a time. The duration of this Go signal will depend on the traffic status determined by the system. Results showed that the system has an average accuracy rate of 80% in vehicle detection and can correctly classify traffic status from the calculated traffic density. [ABSTRACT FROM AUTHOR]

In order to optimize the recognition accuracy of maize kernels breakage detection and improve the detection efficiency of maize kernels breakage, this paper using computer vision technology and detecting of the maize kernels breakage based on K-means clustering algorithm. First, the collected RGB images are converted into Lab images, then the original images clarity evaluation are evaluated by the energy function of Sobel 8 gradient. Finally, the detection of maize kernels breakage using different pixel acquisition equipments and different shooting angles. In this paper, the broken maize kernels are identified by the color difference between integrity kernels and broken kernels. The original images clarity evaluation and different shooting angles are taken to verify that the clarity and shooting angles of the images have a direct influence on the feature extraction. The results show that K-means clustering algorithm can distinguish the broken maize kernels effectively. [ABSTRACT FROM AUTHOR]

In this paper, first anew modified classical conjugate gradient CG method is proposed by deriving a new conjugancy coefficient based on three terms direction, the sufficient descent and the global conjugancy properties are proved. Second: One of the flock algorithms was hybridized which is the particle flock PSO algorithm with a modified conjugate gradient method. The modified PSO Algorithm is normalize the initialization the best by modified CG Algorithm. The numerical results on test functions showed the efficiency of the hybrid algorithm on each of PSO and CG original algorithms. Third: the hybrid algorithm was also used in image processing, by which it was able to estimate image parameter distribution represented by Gibbs random distribution. The estimated parameters was also used in the hybrid algorithm in restoring images and the results were very encouraging. [ABSTRACT FROM AUTHOR]

In this paper, a new detection algorithm based on Convolutional Neural Network is presented in order to realize the fast and convenient ID information extraction in multiple scenarios. The algorithm uses the mobile device equipped with Android operating system to locate and extract the ID number; Use the special color distribution of the ID card, select the appropriate channel component; Use the image threshold segmentation, noise processing and morphological processing to take the binary processing for image; At the same time, the image rotation and projection method are used for horizontal correction when image was tilting; Finally, the single character is extracted by the projection method, and recognized by using Convolutional Neural Network. Through test shows that, A single ID number image from the extraction to the identification time is about 80ms, the accuracy rate is about 99%, It can be applied to the actual production and living environment. [ABSTRACT FROM AUTHOR]

A good dictionary has direct impact to the result of super-resolution image reconstruction. For solving the problem that dictionary learning only contains representation ability but no class information using K-SVD algorithm,this paper proposes single image super-resolution algorithm based on LC-KSVD (Label consist K-SVD).The algorithm adds classifier parameter constraints into the process of dictionary learning and classifier parameters in the process, making the dictionary possess good representation and discrimination ability. The experimental results show that the algorithm has high reconstruction results and good robustness. [ABSTRACT FROM AUTHOR]

The problem of the automatic image calibration is considered in this paper. The most challenging task of the automatic calibration is a proper detection of the calibration object. The solving of this problem required the appliance of the methods and algorithms of the digital image processing, such as morphology, filtering, edge detection, shape approximation. The step-by-step process of the development of the algorithm and its adopting to the specific conditions of the log cuts in the image's background is presented. Testing of the automatic calibration module was carrying out under the conditions of the production process of the logging enterprise. Through the tests the average possibility of the automatic isolating of the calibration object is 86.1% in the absence of the type 1 errors. The algorithm was implemented in the automatic calibration module within the mobile software for the log deck volume measurement. [ABSTRACT FROM AUTHOR]

A relative distance-oriented phase unwrapping algorithm is presented in this paper. Considered the wrapped phase value and modulation distribution of the neighboring pixel, a relative distance of two adjacent pixels is calculated and localized in a complex coordinates, in which the relative distance tree will be composed by all the relative distances and will be used to determine the optimized path of phase unwrapping. The smaller relative distance shows that the phase difference between the two corresponding pixels is very small and the phase data waiting for unwrapping is more reliable. The closer the relative distance is, the more success the phase unwrapping will achieve. Combining the minimum spanning tree algorithm, the phase unwrapping order of each pixel can be determined and the whole phase unwrapping path can also be given. The phase unwrapping path is always directed from the minimum distance value to the greater one. Consequently, this algorithm could avoid error propagating in the phase unwrapping. The errors could be limited in the minimum local region in worst case and the error probability of the phase unwrapping is as low as possible. The physical significance of the relative distance and the fully algorithm for phase unwrapping are proposed in this paper. The result of the experiment show that this new algorithm is feasible and effective, it could control the path avoid crossing the poles, branch-cut and the shadow in the phase unwrapping. [ABSTRACT FROM AUTHOR]

Due to the importance of the two-phase flow researches for the industrial safety analysis, many researchers developed various methods and techniques to study the two-phase flow phenomena on the industrial cases, such as in the chemical, petroleum and nuclear industries cases. One of the developing methods and techniques is image processing technique. This technique is widely used in the two-phase flow researches due to the non-intrusive capability to process a lot of visualization data which are contain many complexities. Moreover, this technique allows to capture direct-visual information data of the flow which are difficult to be captured by other methods and techniques. The main objective of this paper is to present an improved algorithm of image processing technique from the preceding algorithm for the stratified flow cases. The present algorithm can measure the film thickness (hL) of stratified flow as well as the geometrical properties of the interfacial waves with lower processing time and random-access memory (RAM) usage than the preceding algorithm. Also, the measurement results are aimed to develop a high quality database of stratified flow which is scanty. In the present work, the measurement results had a satisfactory agreement with the previous works. [ABSTRACT FROM AUTHOR]

Self Organization Map (SOM) is an automatic tool in data analysis in data mining,it is used to explore the multi-dimentional data which simplifies complexity and produce meaningful relation with each other or high dimentional into low dimentional .The powerful method of SOM i.e learning method results excellent performance .the SOM algorithum have various steps from starting stage to the final neuron and their weight updation and modification ,these procedure resultant a lot of complexity according to the parameters on the basis of experiments, this paper will compare and discuss various parameters and their result or factors that can improve and refine the image through various process of SOM. [ABSTRACT FROM AUTHOR]

Eddy current probes (EC) with Giant Magnetoresistive (GMR) sensors have recently emerged as a promising tool for rapid scanning of multilayer aircraft panels that helps detect cracks under fastener heads. However, analysis of GMR data is challenging due to the complexity of sensed magnetic fields. Further, probes that induce unidirectional currents are insensitive to cracks parallel to the current flow. In this paper, signal processing algorithms are developed for mixing data from two orthogonal EC-GMR scans in order to generate pseudo-rotating electromagnetic field images of fasteners with bottom layer cracks. Finite element simulations demonstrate that the normal component of numerically computed rotating field has uniform sensitivity to cracks emanating in all radial directions. The concept of pseudo-rotating field imaging is experimentally validated with the help of MAUS bilateral GMR array (Big-MR) designed by Boeing. [ABSTRACT FROM AUTHOR]

We discuss a range of Bayesian modelling approaches for spatial data and investigate some of the associated computational challenges. This paper commences with a brief review of Bayesian mixture models and Markov random fields, with enabling computational algorithms including Markov chain Monte Carlo (MCMC) and integrated nested Laplace approximation (INLA). Following this, we focus on the Potts model as a canonical approach, and discuss the challenge of estimating the inverse temperature parameter that controls the degree of spatial smoothing. We compare three approaches to addressing the doubly intractable nature of the likelihood, namely pseudo-likelihood, path sampling and the exchange algorithm. These techniques are applied to satellite data used to analyse water quality in the Great Barrier Reef. [ABSTRACT FROM AUTHOR]

Indirect Immunofluorescence (IIF) is a good standard used for antinuclear autoantibody (ANA) test using Hep- 2 cells to determine specific diseases. Different classifier algorithm methods have been proposed in previous works however, there still no valid set as a standard to classify the fluorescence intensity. This paper presents the use of fuzzy logic to classify the fluorescence intensity and to determine the positivity of the Hep-2 cell serum samples. The fuzzy algorithm involves the image pre-processing by filtering the noises and smoothen the image, converting the red, green and blue (RGB) color space of images to luminosity layer, chromaticity layer 'a' and 'b' (LAB) color space where the mean value of the lightness and chromaticity layer 'a' was extracted and classified by using fuzzy logic algorithm based on the standard score ranges of antinuclear autoantibody (ANA) fluorescence intensity. Using 100 data sets of positive and intermediate fluorescence intensity for testing the performance measurements, the fuzzy logic obtained an accuracy of intermediate and positive class as 85% and 87% respectively. [ABSTRACT FROM AUTHOR]

The goal of this paper is to present a spectral image processing algorithm that is capable to convert a spectral image into a sRGB image and then to determine the spectral characteristics of the colors which contribute to the image of a colored object. The reflective color measurements of a specific area of the image can be carried out and analyzed, making possible to identify the dyes in that area by using white color as reference. The colored images are generated taking as input data the optical spectra of Rhodamine 6G thin films deposited from solutions with different concentrations. Optical spectroscopic and spectral image processing algorithm investigations of the colored thin films present a gradual attenuation of the light in the visible range, revealing significant filtering properties. [ABSTRACT FROM AUTHOR]

The objective of this paper is to present a comprehensive approach to image processing for Giant Magnetoresistive (GMR) array sensors with sheet current sources for the inspection of fatigue cracks in complex metallic structures. The approach incorporates optimal phase angle adjustment, array sensitivity compensation, fastener site identification, an adaptive edge evaluation and removal algorithm and a classification metric. Processing examples are presented that highlight the benefits of these algorithms for improving crack detection for challenging edge and fastener spacing conditions. [ABSTRACT FROM AUTHOR]

The paper presents a software implementation of an Intelligent System for Radiogram Analysis (ISAR). The system has to support radiologists in welds quality inspection. The image processing part of software with a graphical user interface and a welds classification part are described with selected classification results. Classification was based on a few algorithms: an artificial neural network, a k-means clustering, a simplified k-means and a rough sets theory. [ABSTRACT FROM AUTHOR]

In this paper we propose an hybrid method for the image segmentation which combines the edge-based, region-based and the morphological techniques in conjunction through the spectral based clustering approach. An initial partitioning of the image into atomic regions is set by applying a watershed method to the image gradient magnitude. This initial partition is the input to a computationally efficient region segmentation process which produces the final segmentation. We have applied our approach on several images of the Berkeley Segmentation Dataset. The results reveal the accuracy of the propose method. [ABSTRACT FROM AUTHOR]

In this paper we present a concept for an Intelligent System for Radiogram Analysis (ISAR) for welds quality inspection. Both, hardware and software solutions have been introduced in the system. The software operates with variety of scanner standards. It contains preliminary image processing (linear and nonlinear filtering algorithms) and some specialized functions, like Sauvola's tresholding or IQI detection. The aim of the ISAR system is to support a radiologist in his work. [ABSTRACT FROM AUTHOR]

A 'Compton' detector finds the direction of an X-ray by letting it interact with a gaseous, liquid or thin solid material (Tracker) and employing no collimators. This paper takes into account the case of an 'incomplete' solid Tracker where the recoiling electron travels only a few dozen microns and cannot be followed. However, impact positions and incoming and outgoing energies are measured. In this situation, exploiting the Compton Scattering formula, one is only able to identify a cone whose surface the X-ray belongs to. On the other hand, Compton tomography luckily requires only a few views (for example rotating the apparatus in just four positions around the subject), as the 'electronic collimation' that takes place in each position already extracts X-rays coming from many directions. A back-projection algorithm that combines the reconstructed 'cones' in space, weighed according to the Klein-Nishina formula, has been applied to the special case of small animal SPECT (Single Photon Emission Computed Tomography). Here the source is close to the detector, every imaged point in space is calculated from many rays that are emitted at different angles, and the algorithm totally differs from that of the Compton imaging in Astronomy. Tomography reconstruction has been validated by employing simulated data, generated using GEANT4 that includes the Doppler Broadening in the energies of scattered photons, which is due to moving non-free electrons. Space resolution has been assessed. [ABSTRACT FROM AUTHOR]

Computer vision and digital image processing comprises varieties of applications, where some of these used in image processing include convolution, edge detection as well as contrast enhancement. This paper analyzes execution time optimization analysis between Sobel and Canny image processing algorithms in terms of moving objects edge detection. Sobel and Canny edge detection algorithms have been described with pseudo code and detailed flow chart and implemented in C and MATLAB respectively on different platforms to evaluate performance and execution time for moving cars. It is shown that Sobel algorithm is very effective in case of moving multiple cars and blurs images with efficient execution time. Moreover, convolution operation of Canny takes 94–95% time of total execution time with thin and smooth but redundant edges. This also makes more robust of Sobel to detect moving cars edges. [ABSTRACT FROM AUTHOR]

Terahertz time domain pulsed NDT inspection of composite materials creates opportunity to achieve signals allowing to identify various defects types very precisely. In this paper algorithms for signal processing of THz B- and C-scans will be presented. Enhanced THz images enable better defects reconstruction and classification results. Proposed algorithms were verified using signals achieved for various types of defects. For this reason a catalogue of possible defects is under preparation. [ABSTRACT FROM AUTHOR]

Vessel Catheterization is a clinical procedure usually performed by a specialist by means of X-ray fluoroscopic guide with contrast-media. In the present paper, we present a simple and efficient algorithm for vessel segmentation which allows vessel separation and extraction from the background (noise and signal coming from other organs). This would reduce the number of projections (X-ray scans) to reconstruct a complete and accurate 3D vascular model and the radiological risk, in particular for the patient. In what follows, the algorithm is described and some preliminary experimental results are reported illustrating the behaviour of the proposed method. [ABSTRACT FROM AUTHOR]

This paper describes a new circuit that minimizes the number of operations needed to execute the Census transform on 256 grey-level images. The proposed architecture is modular and easily scalable. It can elaborate images with different resolutions and allows regular slice-based layouts to be achieved, thus ensuring also a low design complexity. When realized by using a XILINX VIRTEX4 XC4VLX15 device, the circuit proposed here occupies only 936 slices and reaches a running frequency of about 333 MHz. [ABSTRACT FROM AUTHOR]

In this paper, a cells detection method based on active contours is presented. Active contours or snakes are widely use in image segmentation and tracking problems. Traditionally, active contours initialization is performance manually, this method provides an automatic initialization based on cross correlation. Cell tracking based on active contours provide a faster method to performance this task to multitarget problem. External and internal forces on the image are computed to determinate the action area of the active contours, this task is performance by the Vector Field Convolution (VFC) algorithm. [ABSTRACT FROM AUTHOR]

Image reconstruction for electrical capacitance tomography is to retrieve the permittivity distribution of processes with different dielectric materials inside from the measured capacitances outside. It is a typical inverse problem and has been always a challenge for its nonlinearity and ill-posedness. This paper explores the possibility of applying Tikhonov regularization, widely used for ill-posed problems, to the image reconstruction for electrical capacitance tomography. Two methods using different regularization matrices are investigated by using the simulated data, which are the standard Tikhonov regularization and the Tikhonov regularization based on the second order derivative operator. Results show that both the standard Tikhonov regularization and the Tikhonov regularization with the second order derivative operator do not provide reconstructions with qualities good enough. In order to improve this, a hybrid image reconstruction algorithm, which is based on the Tikhonov regularization with the reconstruction from the linear back projection as a constraint, is developed. Numerical results show that this hybrid image reconstruction algorithm takes advantages from both the linear back projection and the Tikhonov regularization and provide reconstructions better than those from the LBP and the Tikhonov regularization. © 2007 American Institute of Physics [ABSTRACT FROM AUTHOR]

This paper presents a regularization method by using different window functions as regularization for electrical capacitance tomography (ECT) image reconstruction. Image reconstruction for ECT is a typical ill-posed inverse problem. Because of the small singular values of the sensitivity matrix, the solution is sensitive to the measurement noise. The proposed method uses the spectral filtering properties of different window functions to make the solution stable by suppressing the noise in measurements. The window functions, such as the Hanning window, the cosine window and so on, are modified for ECT image reconstruction. Simulations with respect to five typical permittivity distributions are carried out. The reconstructions are better and some of the contours are clearer than the results from the Tikhonov regularization. Numerical results show that the feasibility of the image reconstruction algorithm using different window functions as regularization. © 2007 American Institute of Physics [ABSTRACT FROM AUTHOR]

Electrical capacitance tomography (ECT) is a non-invasive measurement technique that estimates the dielectric permittivity distribution of an inhomogeneous object from the boundary potentials at floating electrodes/or mutual capacitances. In this paper, a stochastic inverse technique based on genetic algorithm (GA-ECT) is developed, which is adapted to the two different methods, i.e. potential measurement and capacitance measurement. Numerical simulation results are presented to evaluate the inverse technique both for noise free and noisy data and the results show that quantitative image can be reconstructed not only with the low permittivity contrast but also with the high contrast. Furthermore, the influence of a priori knowledge to image reconstruction is discussed. © 2007 American Institute of Physics [ABSTRACT FROM AUTHOR]

In this paper an implementation of the Second Order Minimization Algorithm for image tracking using GPU hardware is presented. Several important design considerations for the algorithm efficiency are discussed, such as the choice of data representation and the processing techniques which best match with the capabilities of the current graphic processor architectures. Our experiments show high performance while maintaining the same precision and convergence properties. © 2006 American Institute of Physics [ABSTRACT FROM AUTHOR]

In this paper we present a novel codesign environment which is conceived especially for computer vision hybrid systems. This setting is based on Mathworks Simulink and Xilinx System Generator tools and is comprised of the following: an incremental codesign flow, diverse libraries of virtual components with three levels of description (high level, hardware and software), semi-automatic tools to help in the partition of the system and a methodology for building new library components. The use of high level libraries allows for the development of systems without the need of exhaustive knowledge of the actual architecture or special skills on hardware description languages. This enable a non-traumatic incorporation of the reconfigurable technologies in the image processing systems generally developed for engineers which are not very related to hardware design disciplines. © 2006 American Institute of Physics [ABSTRACT FROM AUTHOR]

The task of recovering the three-dimensional structure of a scene by images of that scene alone is known as the reconstruction problem. To solve this problem it is crucial to understand the geometry of multiple views and to implement the related algorithms. In the past years, many advanced algorithms were developed and published by the computer vision community. As a result, understanding the whole topic well enough to implement the underlying algorithms may mean a prohibitive, time consuming task. This is relevant especially for computer vision students, whose primary research interests base on projective or metric reconstruction. Those may want to use them as mere utilities. So, as always in such a situation, it is time to provide a library, easy to use for the beginner, but flexible enough for the expert. Moreover, such a library would provide a gentle introduction to the computational aspects of the topic. This paper presents the design of such a library, that mainly uses one central data structure and a collection of operators affecting it. © 2006 American Institute of Physics [ABSTRACT FROM AUTHOR]

A simple algorithm finds the partition of a data interval optimizing the fitness of a model that represents the underlying signal as constant over the elements of the partition. Using dynamic programming the exponentially large space of partitions of N data points is implicitly but exhaustively searched in time O(N2). This paper also describes an extension to optimal partitions of higher dimensional data spaces, with application to multivariate signal processing, image processing, cluster analysis, density estimation in 3-dimensional redshift surveys, etc. The algorithm finds the exact global optimum, automatically determines the model order (the number of segments), and has a convenient real-time mode. © 2005 American Institute of Physics [ABSTRACT FROM AUTHOR]

In this paper, we focus on data-limited tomographic imaging problems where the underlying linear inverse problem is ill-posed. A typical regularized reconstruction algorithm uses algebraic formulation with a predetermined discretization resolution. If the selected resolution is too low, we may loose useful details of the underlying image and if it is too high, the reconstruction will be unstable and the representation will fit irrelevant features. In this work, two approaches are introduced to address this issue. The first approach is using Mallow’s CL method or generalized cross-validation. For each of the two methods, a joint estimator of regularization parameter and discretization resolution is proposed and their asymptotic optimality is investigated. The second approach is a Bayesian estimator of the model order using a complexity-penalizing prior. Numerical experiments focus on a space imaging application from a set of limited-angle tomographic observations. © 2005 American Institute of Physics [ABSTRACT FROM AUTHOR]