Narrow Results

A new head tracking algorithm for automatically detecting and tracking human heads in complex backgrounds is proposed. By using an elliptical model for the human head, our Maximum Likelihood (ML) head detector can reliably locate human heads in images having complex backgrounds and is relatively insensitive to illumination and rotation of the human heads. Our head detector consists of two channels: the horizontal and the vertical channels. Each channel is implemented by multiscale template matching. Using a hierarchical structure in implementing our head detector, the execution time for detecting the human heads in a 512×512 image is about 0.02 second in a Sparc 20 workstation (not including the time for image acquisition). Based on the ellipse-based ML head detector, we have developed a head tracking method that can monitor the entrance of a person, detect and track the person's head, and then control the stereo cameras to focus their gaze on this person's head. In this method, the ML head detector and the mutually-supported constraint are used to extract the corresponding ellipses in a stereo image pair. To implement a practical and reliable face detection and tracking system, further verification using facial features, such as eyes, mouth and nostrils, may be essential. The 3D position computed from the centers of the two corresponding ellipses is then used for fixation. An active stereo head has been used to perform the experiments and has demonstrated that the proposed approach is feasible and promising for practical uses.

Edges are prominent features in images. The detection and analysis of edges are key issues in image processing, computer vision and pattern recognition. Wavelet provides a powerful tool to analyze the local regularity of signals. Wavelet transform has been successfully applied to the analysis and detection of edges. A great number of wavelet-based edge detection methods have been proposed over the past years. The objective of this paper is to give a brief review of these methods, and encourage the research of this topic. In practice, an image is usually of multistructure edge, the identification of different edges, such as steps, curves and junctions play an important role in pattern recognition. In this paper, more attention is paid on the identification of different types of edges. We present the main idea and the properties of these methods.

Techniques to identify printed and handwritten text in scanned documents differ significantly. In this paper, we address the question of how to discriminate between each type of writing on registration forms. Registration-form documents consist of various type zones, such as printed text, handwriting, table, image, noise, etc., so segmenting the various zones is a challenge. We adopt herein an approach called “multiscale-region projection” to identify printed text and handwriting. An important aspect of our approach is the use of multiscale techniques to segment document images. A new set of projection features extracted from each zone is also proposed. The classification rules are mining and are used to discern printed text and table lines from handwritten text. The proposed system was tested on 11118 samples in two registration-form-image databases. Some possible measures of efficiency are computed, and in each case the proposed approach performs better than traditional methods.

Edge detection is an active and critical topic in the field of image processing, and plays a vital role for some important applications such as image segmentation, pattern classification, object tracking, etc. In this paper, an edge detection approach is proposed using local edge pattern descriptor which possesses multiscale and multiresolution property, and is named varied local edge pattern (VLEP) descriptor. This method contains the following steps: firstly, Gaussian filter is used to smooth the original image. Secondly, the edge strength values, which are used to calculate the edge gradient values and can be obtained by one or more groups of VLEPs. Then, weighted fusion idea is considered when multiple groups of VLEP descriptors are used. Finally, the appropriate threshold is set to perform binarization processing on the gradient version of the image. Experimental results show that the proposed edge detection method achieved better performance than other state-of-the-art edge detection methods.

CNN-based methods have made great progress in single-image rain removal. Most recent methods improve performance by increasing the depth of the network. To fully extract local and global features while reducing inference time, we propose a top-to-down attribute-insensitive multiscale hourglass network for rain streak and raindrop removal. For the rain removal task, we expect that the constructed network can accurately identify the various attributes of the rain information characteristics of the small target. Considering the difference in the size, shape, direction and density of rain streak and raindrop, inspired by the performance of hourglass architecture to capture multiscale features in human pose estimation, we introduce an attribute-insensitive hourglass module to recognize the attributes of rain streak and raindrop in a unified framework. This feature extraction module could capture the characteristics of rain streak and raindrop with different attributes. This stacked hourglass blocks down-sample features and then up-samples them back to the original resolution based on discrete wavelet transform and inverse discrete wavelet transform. We perform extensive experiments on five synthetic and real-world de-raining datasets to validate the effectiveness of our proposed network on rain streak and raindrop removal. The qualitative and quantitative results show that our method is suitable for removing rain streak and raindrop in a unified framework. We present the results of generalization and ablation study for key components, we also report the accuracy of semantic segmentation after preprocessing with all rain removal methods. Our source code will be available on the GitHub: https://github.com/Ruini94/AIMHNet.

Bone is a multiscale combination of collagen molecules merged with mineral crystals. Its high rigidity and stability stem amply from its polymeric organic matrix and secondly from the connections established between interdifferent and intradifferent scale components through cross-links. Several studies have shown that the cross-links inhibition results in a reduction in strength of bone but they do not quantify the degree to which these connections contribute to the bone rigidity and toughness. This report is classified among the few works that measure the cross-links multiscale impact on the ultrastructure bone mechanical behavior.

This work aims firstly to study the effect of cross-links at the molecule scale and secondly to gather from literature studies results handling with cross-links effects on the other bone ultrastructure scales in order to reveal the multiscale effect of cross-links. This study proves that cross-links increasing number improves the mechanical performance of each scale of bone ultrastructure. On the other hand, cross-links have a multiscale contribution that depends on its rank related to existing cross-links connecting the same geometries and it depends on mechanical characteristics of geometries connected.

Interface and visualization tools usually provide static representations of biological pathways, which can be a severe limitation: fixed pathway boundaries are used without consensus about the elements that should be included in a particular pathway; one cannot generate new pathways or produce selective views of existing pathways. Also, the tools are not capable of integrating multiple levels that conceptually can be distinguished in biological systems.

We present ReConn, an interface and visualization tool for a flexible analysis of large data at multiple biological levels. ReConn (Reactome Connector) is an open source extension to Cytoscape which allows user friendly interaction with the Reactome database. ReConn can use both predefined Reactome pathways as well as generate new pathways. A pathway can be derived by starting from any given metabolite and existing pathways can be extended by adding related reactions. The tool can also retrieve alternative routes between elements of a biological network. Such an option is potentially applicable in the design and analysis of knockout experiments. ReConn displays information about multiple levels of the system in one view. With these dynamic features ReConn addresses all of the above mentioned limitations of the interface tools.

Modeling elastic waves in complex media, with varying physical properties, require very accurate algorithms and a great computational effort to avoid nonphysical effects. Among the numerical methods the spectral elements (SEM) have a high precision and ease in modeling such problems and the physical domains can be discretized using very coarse meshes with elements of constant properties. In many cases, models with very complex geometries and small heterogeneities, shorter than the minimum wavelength, require grid resolution down to the thinnest scales, resulting in an extremely large problem size and greatly reducing accuracy and computational efficiency. In this paper, a poly-grid method (PG-CSEM) is presented that can overcome this limitation. To accurately deal with continuous variations or even small-scale fluctuations in elastic properties, temporary auxiliary grids are introduced that prevent the need to use large meshes, while at the macroscopic level wave propagation is solved maintaining the SEM accuracy and computational efficiency as confirmed by the numerical results.

The magnetic resonance imaging technique is mostly used for visualizing and detecting brain tumor, which requires accurate segmentation of brain MR images into white matter, gray matter, cerebrospinal fluid, necrotic tissue, tumor, and edema. But brain image segmentation is a challenging task because of unknown noise and intensity inhomogeneity in brain MR images. This paper proposed a technique for the segmentation and the detection of a tumor, cystic component and edema in brain MR images using multiscale intuitionistic fuzzy roughness (MSIFR). Application of linear scale-space theory and intuitionistic fuzzy image representation deals with noise and intensity inhomogeneity in brain MR images. Intuitionistic fuzzy roughness calculated at proper scale is used to find optimum valley points for segmentation of brain MR images. The algorithm is applied to the real brain MR images from various hospitals and also to the benchmark set of the synthetic MR images from brainweb. The algorithm segments synthetic brain MR image into three regions, gray matter (GM), white matter (WM) and cerebrospinal fluid (CSF) and also separates tumor, cystic component and edema accurately in real brain MR images. The results of segmentation of proposed algorithm for synthetic images are compared with nonlocal fuzzy c-means (NLFCM), rough set based algorithms, intervalued possibilistic fuzzy c-means (IPFCM), robust modified Gaussian mixture model with rough set (RMGMMRS) and three algorithms, recursive bias corrected possibilistic fuzzy c-means (RBCPFCM), recursive bias corrected possibilistic neighborhood fuzzy c-means (RBCPNFCM) and recursive bias corrected separately weighted possibilistic neighborhood fuzzy c-means (RBCSPNFCM). The quantitative and qualitative evaluation demonstrates the superiority of the proposed algorithm.

Nowadays, the number of patients with brain tumors is steadily increasing, diagnosis and isolation of the tumor play an important role in the process of treatment and surgery. Due to the high error of manual segmentation of the tumor, algorithms that perform this operation with less error are of great importance. Convolutional neural networks have made great progress in the field of medical imaging. The use of imaging techniques and pattern recognition in the diagnosis and automatic determination of brain tumors by MRI imaging reduces errors, human error and speeds up detection. The artificial convolutional neural network (CNN) has been widely used in the diagnosis of intelligent cancers and has significantly reduced the error rate. Therefore, in this paper, we present a new method using a combination of convolutional and multi-scale artificial neural network that has significantly increased the accuracy of tumor diagnosis. This study presents a multidisciplinary convolution neural network (MCNN) approach to classifying tumors that can be used as an important part of automated diagnosis systems for accurate cancer diagnosis. Based on the MCNN structure, which presents the MRI image to several deep convolutional neural networks of varying sizes and resolutions, the stage of extracting classical hand-made features is avoided. This approach proposes better classification rates than the classical methods. This study uses a multi-scale convolution technique to achieve a detection accuracy of 95/4%, which shows the efficiency of the proposed method.

The parallel, unstructured-mesh SWAN model has been employed to study a tidal inlet with complex bathymetry in the Dutch Wadden Sea. The unstructured grid resolves the large-scale, O(1km) wave dynamics in the open sea whereas employing 15-20m grid-resolution over the tidal basins and flats to gain an insight of surf breaking, local wind-wave and wave-current interactions and to assess the performance of SWAN through hindcasts of storm events. Previous studies and the present study show that the default source term settings routinely underestimate the low-frequency peak energy in the wind-sea part of the spectrum and the finite-depth wave growth over nearly horizontal beds. High-resolution simulations have indicated that the accuracy is directly affected by parameterization of bottom friction, whitecapping dissipation and depth-induced wave breaking.