Markov random field based automatic image alignment for electron tomography

We propose a self-learning method for automatic wafer alignment in the semiconductor manufacturing process. A feed forward neural network is trained by and used for wafer alignment. The network determines the movement of kinematic parts from the misalignment inspected by machine vision. However, it is time-consuming and inconvenient to obtain training data in this way. So, we built an automatic learning rule to gather the data and train the network. The network may determine wrong outputs and cause other misalignments at first, but the error can decrease as the training proceeds. The training sets consisted of a variation of misalignment data and the movement of an alignment stage. Five recent sets are used for training and others are dismissed or forgotten. This retrained network tried aligning, measured misalignment, and made new training sets. This sequence makes it possible to acquire alignment skill and automate the process. After learning, automatic alignment accomplished sub-pixel accuracy for several cases of misalignment. The result showed that the proposed method could be applied to the semiconductor manufacturing process. Its performance improved about 6% compared with conventional algorithms.

This paper presents a new method to automatically align specific scale-marks for the calibration of hydrometers. A hydrometer calibration system adopting the new method consists of a vision system, a stepping motor, and software to control the system. The vision system is composed of a CCD camera and a frame grabber, and is used to acquire images. The stepping motor moves the camera, which is attached to the vessel containing a reference liquid, along the hydrometer. The operating program has two main functions: to process images from the camera to find the position of the horizontal plane and to control the stepping motor for the alignment of the horizontal plane with a particular scale-mark. Any system adopting this automatic alignment method is a convenient and precise means of calibrating a hydrometer. The performance of the proposed method is illustrated by comparing the calibration results using the automatic alignment method with those obtained using the manual method.

Explicit fusion of perfusion data from Positron Emission Tomography (PET) or Single Photon Emission Computed Tomography (SPECT) with coronary artery anatomy from Computed Tomographic Coronary Angiography (CTA) has been shown to improve the diagnostic yield for coronary artery disease (CAD) compared to either modality alone. However, most clinically available methods were developed for multimodal scanners or require interactive alignment prior to display and analysis. A new approach was developed to register the two distributions obtained either from a single multimodal imager or from separate scanners, and a preliminary validation was undertaken to compare the automatic alignment to interactive alignment by two experts.

Automatic alignment (matching) of two-dimensional gel electrophoresis images is of primary interest in the evolving field of proteomics. In the present study, feature-based matching techniques, in their classical and robust versions, are described, and an automatic method of fuzzy alignment (FA) is introduced. This method allows automatic matching of two gel images with different numbers of features with unknown correspondence. Performance of FA is tested on simulated and real data sets.

Local feature based methods have achieved a great success in the field of image matching due to its invariance under typical image transformations. However, local features are often not invariant under complex non-affine transformations, which makes the matching methods ineffective. To remove the effect of complex image transformations, this paper proposes alignment of images before they are compared. An automatic image alignment method is introduced based on thin plate spline (TPS) warping. Then, the aligned matching scheme is designed to utilize correct alignments and reject false alignments for the improvement of matching performance. The method is evaluated using scene retrieval and object categorization. Experiments show the proposed aligned matching outperforms two typical methods: voting scheme and histograms comparison (over a set of prototypes which must be found by clustering).

This article deals with the issue of the automatic setting of machines on their base. The focus is on machines with a self-supporting frame, freestanding, on more than three supports. The alignment uses the previously presented method of precise and repeatable machine placement on the foundation. As a result, the machine has the necessary parameters in terms of its geometry, and at the same time a defined weight distribution on the individual building elements is guaranteed. The paper presents a new method for automatic machine alignment using so-called “smart feet”. Due to this, the process of machine alignment is clearly defined, repeatable and does not depend on the experience of the personnel performing the alignment. The presented algorithm makes it possible to automate the entire machine-setting process. The article includes the presentation of a machine placement on its base, which is used for testing and verification of the functionality of the alignment algorithm.

The Semantic Web is based on technologies that make the content of the Web machine-understandable. In that framework, ontological knowledge representation has become an important tool for the analysis and understanding of multimedia information. Because of the distributed nature of the Semantic Web however, ontologies describing similar fields of knowledge are being developed and the data coming from similar but non-identical ontologies can be combined only if a semantic mapping between them is first established. This has lead to the development of several ontology alignment tools. We propose an automatic ontology alignment method based on the recursive neural network model that uses ontology instances to learn similarities between ontology concepts. Recursive neural networks are an extension of common neural networks, designed to process efficiently structured data. Since ontologies are a structured data representation, the model is inherently suitable for use with ontologies.

This article presents a Markov Random Field (MRF) coupled energy function designed for non rigid image registration. The proposed function addresses the difficulties associated with accurate image registration while maintaining spatial coherence. It combines intensity based and feature based terms. The intensity based term measures the similarity between pixels in fixed and moved images for registration, while the feature based term uses the Scale-Invariant Feature Transform (SIFT) method, ensuring robust alignment of pixel intensities and structural features between image pairs. Two graph cut optimization algorithms, α-expansion, and αβ-swap, are evaluated to optimize the coupled energy function proposed. Experimental results calculated for all pairs in the FIRE retinal dataset demonstrate that the α-expansion algorithm outperforms the αβ-swap algorithm. The α-expansion algorithm exhibits a lower mean squared error (MSE) indicating that it creates a registered image more similar to the fixed in terms of similarity, and a higher structural similarity index value (SSIM), indicating that the algorithm preserves the important visual features of the image. Furthermore, results confirm that the proposed coupled MRF based registration approach significantly outperforms the traditional intensity based registration method.

Although diffusion tensor imaging (DTI) has been widely used for the quantitative analyses of the integrity of white matter in the brain in clinical and research fields, quality assurance (QA) for DTI has not been fully established. Thus, we suggest a QA guideline for DTI using the American College of Radiology (ACR) Magnetic resonance imaging (MRI) head phantom. In this study, the geometric accuracy, slice-position accuracy, image intensity uniformity, percent signalghosting, low-contrast object detectability, image distortion, fractional anisotropy (FA), and apparent diffusion coefficient (ADC) were measured and evaluated in 1.5T and 3.0T MRI scanners equipped with an 8-channel SENSE head coil. The standard axial spin echo (SE) T1-weighted MR images and DTI with 6, 15 and 32 directions were obtained. Concerning geometric accuracy, image twisting in the three directions was observed due to the inhomogeneity of echo planar imaging (EPI). Image intensity uniformity was significantly lower for DTI than for the standard SE T1-weighted MR images. Percent signal ghosting was higher for images from 3.0T MRI than for images from 1.5T MRI. Low-contrast object detectability was visually identified and measured at a low contrasttonoise ratio (CNR) and a low signaltonoise ratio (SNR). Image distortion changed remarkably to the phaseencoding direction. The present study using the ACR MRI phantom suggests a QA method for DTI with high reproducibility and easy accessibility.

XAOM: A method for automatic alignment and orientation of radiographs for computer-aided medical diagnosis

Graph and hypergraph matching are fundamental problems in computer vision, which have been receiving a steadily growing attention in the last two decades due to the recent modern algorithms and models that are both fast and accurate. Graph and hypergraph matching are successful in tackling many real-world tasks that require 2D and 3D feature matching and alignment. Matching is a general problem in vision and graphs have always been quintessential for modeling information that has both local and longer range connections. Therefore, graph matching will continue to influence computer vision and robotics research, crossing over many different mathematical and computational models, including Bayesian networks, conditional and Markov random fields, spectral graph models, and, more recently, deep neural networks. While graph matching is limited to using pairwise relationships, hypergraph matching permits the use of relationships between sets of features of any order. Consequently, it carries the promise to make matching more robust to changes in scale, deformations, and outliers. In this chapter, we present several key ideas, computational models, and optimization methods that introduced in the literature the possibility to learn graph and hypergraph matching in a semi-supervised or even completely unsupervised fashion. We start by describing two methods for solving graph and hypergraph matching based on spectral matching and integer projected fixed point methods, introduced in Chap. 1. Then, based on the same intuition (also discussed in Chap. 1) that correct assignments form a strong cluster of agreements, we follow up by introducing the main approach to learning as it applies to graph and hypergraph matching, as well as the more general Maximum A Posteriori (MAP) inference problem in probabilistic graphical models (MRFs and CRFs). In the last part of the chapter, we demonstrate experimentally the effectiveness of the algorithms presented and, whenever possible, provide extensive comparisons to other top published methods. Moreover, we show that our learning approach is general and can significantly boost the performance of different state-of-the-art matching algorithms in the literature.

Nanofabrication of a gold fiducial array on specimen support for electron tomography

Raster image vectorization is increasingly important since vector-based graphical contents have been adopted in personal computers and on the Internet. In this paper, we introduce an effective vector-based representation and its associated vectorization algorithm for full-color raster images. There are two important characteristics of our representation. First, the image plane is decomposed into nonoverlapping parametric triangular patches with curved boundaries. Such a simplicial layout supports a flexible topology and facilitates adaptive patch distribution. Second, a subset of the curved patch boundaries are dedicated to faithfully representing curvilinear features. They are automatically aligned with the features. Because of this, patches are expected to have moderate internal variations that can be well approximated using smooth functions. We have developed effective techniques for patch boundary optimization and patch color fitting to accurately and compactly approximate raster images with both smooth variations and curvilinear features. A real-time GPU-accelerated parallel algorithm based on recursive patch subdivision has also been developed for rasterizing a vectorized image. Experiments and comparisons indicate our image vectorization algorithm achieves a more accurate and compact vector-based representation than existing ones do. © 2009 ACM.

This study presents an automatic visibility retrieval of a FLIR A320 Stationary Thermal Imager installed on a measurement tower on the mountain Lagern located in the Swiss Jura Mountains. Our visibility retrieval makes use of edges that are automatically detected from thermal camera images. Predefined target regions, such as mountain silhouettes or buildings with high thermal differences to the surroundings, are used to derive the maximum visibility distance that is detectable in the image. To allow a stable, automatic processing, our procedure additionally removes noise in the image and includes automatic image alignment to correct small shifts of the camera. We present a detailed analysis of visibility derived from more than 24000 thermal images of the years 2015 and 2016 by comparing them to (1) visibility derived from a panoramic camera image (VISrange), (2) measurements of a forward-scatter visibility meter (Vaisala FD12 working in the NIR spectra), and (3) modeled visibility values using the Thermal Range Model TRM4. Atmospheric conditions, mainly water vapor from European Center for Medium Weather Forecast (ECMWF), were considered to calculate the extinction coefficients using MODTRAN. The automatic visibility retrieval based on FLIR A320 images is often in good agreement with the retrieval from the systems working in different spectral ranges. However, some significant differences were detected as well, depending on weather conditions, thermal differences of the monitored landscape, and defined target size.

Domain adaptive object detection (DAOD) aims to generalize an object detector trained on labeled source-domain data to a target domain without annotations, the core principle of which is source-target feature alignment. Typically, existing approaches employ adversarial learning to align the distributions of the source and target domains as a whole, barely considering the varying significance of distinct regions, say instances under different circumstances and foreground vs background areas, during feature alignment. To overcome the shortcoming, we investigate a differential feature alignment strategy. Specifically, a prediction-discrepancy feedback instance alignment module (dubbed PDFA) is designed to adaptively assign higher weights to instances of higher teacher-student detection discrepancy, effectively handling heavier domain-specific information. Additionally, an uncertainty-based foreground-oriented image alignment module (UFOA) is proposed to explicitly guide the model to focus more on regions of interest. Extensive experiments on widely-used DAOD datasets together with ablation studies are conducted to demonstrate the efficacy of our proposed method and reveal its superiority over other SOTA alternatives.