Operator-Norm-Based Variable-Wise Diagonal Preconditioning for Automatic Stepsize Selection of A Primal-Dual Splitting Algorithm

Object-constraint programming provides a design to integrate constraints with dynamic, object-oriented programming languages. It allows developers to encode multi-way constraints over objects and object collections using existing, object-oriented abstractions. These constraints are automatically maintained at run-time. One original goal of the Babelsberg-family of object-constraint programming languages was to allow users familiar with the imperative paradigm to quickly and efficiently make use of constraint solver capabilities. Yet, practical problems often require careful selection of solvers to find good solutions (or any at all). Furthermore, solver performance can vary and while most solvers come with various optimizations, developers have to have a good understanding of the solving process to use these optimizations effectively. This, however, is difficult to achieve if the solver is automatically selected by the system. In this work, we discuss three different implementations for automatic solver selection that we used in Babelsberg implementations. As a second step, we look at the performance potential of edit constraints that are available in some solvers such as Cassowary or DeltaBlue, and how they can be applied automatically to improve solver performance. We argue that these techniques make object-constraint programming more practical by improving the quality and performance of solutions.

In order to give a virtual human rich and realistic facial expression in the film production process, a good blendshape model is needed. But selecting and capturing base expressions for blendshape model requires a lot of manual work, time and effort, and the model also lacks expressiveness. A method for automatically selecting a set of base expressions from a sequence of facial motions is proposed in this paper. In this method, the Procrustes analysis is used to estimate the difference between face meshes and determine the composition of the base expressions. And the base expressions are used to build a local blendshape model which can enhance expressiveness. The results of reconstructing facial expressions by the local blendshape model are shown in this paper. By this method, the base expressions can be automatically selected from the expression sequence, reducing the manual operation.

In this paper, we investigate the problem of automatically selecting textual keywords for keyword search development and tuning on audio data for any language. Briefly, the method samples candidate keywords in the training data while trying to match a set of target marginal distributions for keyword features such as keyword frequency in the training or development audio, keyword length, frequency of out-of-vocabulary words, and TF-IDF scores. The method is evaluated on four IARPA Babel program base period languages. We show the use of the automatically selected keywords for the keyword search system development and tuning. We show also that search performance is improved by tuning the decision threshold on the automatically selected keywords.

We present a method for automatically selecting the best filter to treat poorly printed documents using image quality assessment. In order to estimate the quality of the image, we introduce five quality measures: stroke thickness factor, broken character factor, touching character factor, small speckle factor, and white speckle factor. Based on the information provided by the quality measures, a set of rules uses a two-stage decision process to choose the best filter among 4 morphological filters to be applied to an image. Other preprocessing tasks implemented are: skew correction, connected components analysis, and detection of reference lines. Our database contains 736 document images that were divided in three sets: training, validation and testing. Most images have one or more of the following degradations: broken characters, touching characters and salt-and-pepper noise. A training set of 370 images was used to develop the system. Experimental results on the test set of 183 images show a significant improvement in the recognition rate from 73.24% using no filter at all to 93.09% after applying a filter that was automatically selected. The recognition rate refers to the number of characters that were correctly recognized in the image using a commercial OCR. Three commercial OCR's were used to demonstrate the improvement obtained in the recognition rates in the training set.

The treatment of thoracic cancer using external beam radiation requires an optimal selection of the radiation beam directions to ensure effective coverage of the target volume and to avoid unnecessary treatment of normal healthy tissues. Intensity modulated radiation therapy (IMRT) planning is a lengthy process, which requires the planner to iterate between choosing beam angles, specifying dose-volume objectives and executing IMRT optimization. In thorax treatment planning, where there are no class solutions for beam placement, beam angle selection is performed manually, based on the planner's clinical experience. The purpose of this work is to propose and study a computationally efficient framework that utilizes machine learning to automatically select treatment beam angles. Such a framework may be helpful for reducing the overall planning workload. The authors introduce an automated beam selection method, based on learning the relationships between beam angles and anatomical features. Using a large set of clinically approved IMRT plans, a random forest regression algorithm is trained to map a multitude of anatomical features into an individual beam score. An optimization scheme is then built to select and adjust the beam angles, considering the learned interbeam dependencies. The validity and quality of the automatically selected beams evaluated using the manually selected beams from the corresponding clinical plans as the ground truth. The analysis included 149 clinically approved thoracic IMRT plans. For a randomly selected test subset of 27 plans, IMRT plans were generated using automatically selected beams and compared to the clinical plans. The comparison of the predicted and the clinical beam angles demonstrated a good average correspondence between the two (angular distance 16.8° ± 10°, correlation 0.75 ± 0.2). The dose distributions of the semiautomatic and clinical plans were equivalent in terms of primary target volume coverage and organ at risk sparing and were superior over plans produced with fixed sets of common beam angles. The great majority of the automatic plans (93%) were approved as clinically acceptable by three radiation therapy specialists. The results demonstrated the feasibility of utilizing a learning-based approach for automatic selection of beam angles in thoracic IMRT planning. The proposed method may assist in reducing the manual planning workload, while sustaining plan quality.

A load-adaptive automatic switching frequency selection scheme is proposed to improve the power efficiency of a switching buck converter at light load condition. The buck converter operates in the continuous-conduction mode for heavy loading and the switching frequency is fixed at its maximum value. For light loading, the buck converter operates in the discontinuous-conduction mode and its switching frequency is automatically selected among a pre-defined set of frequencies according to the amount of the load current. The load current can be sensed indirectly by monitoring the on-time of power transistor because it is a function of the load current. With the proposed load-adaptive automatic switching frequency selection circuit, the power efficiency of a buck converter implemented in a 0.35-μm 2P4M BCDMOS technology is improved by 24.0-% when the load current load is 10-mA.

To simplify the often complex and user-dependent manual region of interest (ROI) selection process for head motion monitoring, an automatic ROI selection method was developed. The automatic ROI selection algorithm calculated the displacements and velocities of 3D surface points between a temporally correlated 3D image series and a reference image. Only facial surfaces satisfying certain spatial and temporal criteria were selected. The algorithm was tested on five healthy volunteers instructed to perform different types of facial movements for a total of 27 real-time image sets (40-120 images for each image set). The algorithm detected and excluded surface areas affected by different types of local facial movements that were independent of actual net head motion. Eye, eyebrow, and mandible motion were most commonly detected as being independent of head motion and were excluded from the final ROI. For 3D images taken with substantial facial or whole head motion, either most of the facial area was excluded or only small areas with random patterns were included in the final ROI. Surface image registration using iterative closest point (ICP) methods showed more stable real-time head tracking using the automatically selected ROI than manual user defined ROIs. The automatic selection method successfully found ROIs stable over time for tracking head motion by excluding locally varying facial motions. By automating the ROI selection process, it is feasible that the time and complexity of current ROI definition can be reduced, together with user-dependent registration errors.

This paper presents a single-inductor triple-source quad-mode (SITSQM) energy-harvesting interface in a 0.18- $\mu \text{m}$ CMOS process. The proposed reversely polarized energy recycling (RPER) technique improves not only the conversion efficiency at low input voltage but also the system’s output power range. The interface employs the buck–boost topology to convert energy from photovoltaic (PV) cells and a thermoelectric generator (TEG) to a regulated 1.2-V output. The proposed converter features four different operating modes, namely, harvesting, recycling, storing, and backup. The operating mode is automatically selected according to the input and load conditions using the automatic source selection mechanism. The experimental results demonstrate 25.3% efficiency improvement and 10 $\times $ output power range extension from the proposed RPER technique. The proposed SITSQM converter automatically manages three harvesting sources with 82.1% peak conversion efficiency.

On the relevance of automatically selected single-voxel MRS and multimodal MRI and MRSI features for brain tumour differentiation

Video super-resolution based on automatic key-frame selection and feature-guided variational optical flow

BackgroundReduced image quality from increased X-ray scatter and image noise can be problematic when coronary computed tomography angiography (CCTA) imaging is performed in obese patients. The aim of this study was to compare the image quality and radiation dose obtained using automatic tube potential selection with tube current modulation (APSCM) with those obtained using a body mass index (BMI)-based protocol for CCTA in obese patients.MethodsA total of 203 consecutive obese (BMI > 30 kg/m2) patients were retrospectively enrolled, of whom 96 underwent CCTA with APSCM and 107 underwent a BMI-based protocol. After applying the propensity score matching method, the clinical parameters, subjective and objective image quality, and radiation dose were compared between the APSCM group and the matched BMI-based group. These parameters were also compared among different tube potential subgroups.ResultsNo significant differences were observed between the APSCM group and the BMI-based group with respect to image quality or radiation dose assessment (p > 0.05). Twenty patients (21%) examined with 140 kV in the APSCM group were exposed to significantly more radiation (p < 0.05) than patients in the BMI-based group or patients in the other APSCM kV subgroups; significant improvement in image quality was not observed in the 140 kV subgroup. Patients with a high BMI and a large effective diameter tended to be examined with 140 kV (p < 0.05).ConclusionThe use of APSCM for CCTA in obese patients did not significantly reduce the radiation dose or improve image quality compared with those in the matched BMI-based group. Our data indicate that it is better to avoid using APSCM when 140 kV is automatically selected, due to increased radiation dose and lack of significant improvement in image quality.

The study deals with automatic polishing performed by a six-degree-of-freedom industrial robot equipped with two types of automatically selected rotational polishing tools. Little progress has been made in the autolmation of polishing since it requires much skill and experience. An industrial robot and a CAD system are introduced to cope with this problem. However, most products are so complicated that one type of polishing tool is not enough to provide a good finish. The aim of this study is to device a program that will enable the robot to automatically select polishing tools suitable for each workpiece surface and to polish it according to the type of tool. The system automatically selects the I-shaped (straight) rotational tool or the L-shaped one according to the surface shape of a workpiece using strategies to discriminate on the basis of CAD data. After the selection, the system generates a polishing path for both tools and performs the polishing. From experimental results, the system is found to be effective in polishing workpiece surfaces of complicated shapes.

Recently, laser scanning systems, onboard airborne and terrestrial mobile mapping systems, have been established as a leading technology for collecting high density 3D information from an object's surface. The availability of generated surface models is very important for various industrial, military, environmental, and public applications. The accuracy of the derived point cloud coordinates from a LiDAR system is affected by inherent systematic and random errors. The impact of random errors depends on the precision of the system’s measurements, which comprise position and orientation information from the GPS/INS unit, mirror angles, and ranges. On the other hand, systematic errors are mainly caused by biases in the mounting parameters (i.e., lever arm offset and boresight angles) relating the system components as well as biases in the system measurements (e.g., ranges and mirror angles). In order to ensure the geometric quality of the collected point cloud, the LiDAR systems should undergo a rigorous calibration procedure to estimate the system parameters that minimize the discrepancies between conjugate surface elements in overlapping LiDAR strips. The main objective of this paper is to look into an existing LiDAR system calibration technique, which is based on manual selection of overlapping regions between LiDAR strips and how to increase the efficiency of this technique by automatic selection of appropriate overlapping strip pairs, which should achieve the minimum optimal flight configuration that maximizes the impact of the discrepancies among conjugate surface elements in overlapping strips as well as automatic selection of regions within the appropriate overlapping strip pairs. The methodology of the proposed technique can be summarized as follows: first, the LiDAR strip pairs are grouped based on the flight configuration; second, appropriate overlapping strip pairs from each group is automatically selected; third, regions within the appropriate overlapping strip pairs are automatically selected based on their angles (slopes and aspects) and distribution; finally, the calibration procedure is applied. The experimental results have shown that the quality of the estimated parameters using the automatic selection are quite comparable to the estimated parameters using the manual selection while the proposed method is fully automated, and much faster.

Scale-spaces induced by diffusion processes play an important role in many computer vision tasks. Automatically selecting the most appropriate scale for a particular problem is a central issue for the practical applicability of such scale-space techniques. This paper concentrates on automatic scale selection when nonlinear diffusion scale-spaces are utilized for image denoising. The problem is studied in a statistical model selection framework and cross-validation techniques are utilized to address it in a principled way. The proposed novel algorithms do not require knowledge of the noise variance and have acceptable computational cost. Extensive experiments on natural images show that the proposed methodology leads to robust algorithms, which outperform existing techniques for a wide range of noise types and noise levels.

Autoregressive (AR) models are fundamental tools for modeling a variety of signals in many fields of study. Selecting the appropriate order for AR models is typically accomplished using an information criterion to compare the models learned for all orders under consideration. The use of an information criterion for model order selection becomes increasingly computationally demanding when AR models are used as part of larger statistical models, such as mixture models, that have their own model order selection issues. Statistical models utilizing Dirichlet process (DP) priors provide a mechanism for automatically selecting the number of components within a mixture model, and have previously been utilized with AR components. These previous investigations utilize different priors for the AR parameters to enable automatic selection of the AR order and each makes us of computationally expensive Markov chain Monte Carlo (MCMC) sampling. This paper develops and evaluates a variational Bayesian (VB) inference procedure for the parameters of DP mixtures of AR components with uncertain order, to enable rapid parameter inference for AR based statistical models that provide automatic model order selection and is suitable for large scale problems. The previously utilized priors for AR models with uncertain order are evaluated to determine which is more appropriate for VB inference and the ability of the VB inference procedure for the developed model to correctly determine the number of mixture components and the AR order of each of the mixture components is shown to be comparable to computationally intensive MCMC inference. The VB inference procedure is then applied to an acoustic signal classification problem to illustrate the efficacy of AR based statistical models utilizing automated model order selection for real-world signal processing tasks.