The Infona portal uses cookies, i.e. strings of text saved by a browser on the user's device. The portal can access those files and use them to remember the user's data, such as their chosen settings (screen view, interface language, etc.), or their login data. By using the Infona portal the user accepts automatic saving and using this information for portal operation purposes. More information on the subject can be found in the Privacy Policy and Terms of Service. By closing this window the user confirms that they have read the information on cookie usage, and they accept the privacy policy and the way cookies are used by the portal. You can change the cookie settings in your browser.
One of recent trends [31, 32, 14] in network architecture design is stacking small filters (e.g., 1x1 or 3x3) in the entire network because the stacked small filters is more efficient than a large kernel, given the same computational complexity. However, in the field of semantic segmentation, where we need to perform dense per-pixel prediction, we find that the large kernel (and effective receptive...
The aim of scene semantic segmentation is to label each pixel with a class which it belongs to in high level cognition. State-of-art works mainly adapt convolutional neural networks originally designed for image classification to make dense prediction. However the inner structure of scene itself and its stuff is more flexible and variable, which is distinct from the objects in image classification...
Image segmentation is an important problem in image processing and object recognition, and is one well-known bottleneck for further applications. Fuzzy C-means, as one typical clustering algorithm in pattern recognition, has been improved for image segmentation in many aspects. Aiming at the distance form in FCM, this paper proposes to incorporate FCM with kernel functions, which will make it insensitive...
In this paper, we present an approach to recognize the Machine-printed Mongolian characters by CNN (convolutional neural network). Firstly, a training set of traditional Mongolian characters is collected in advance. There are 85 categories in all and each category is trained and recognized by the CNN. And then, we set a CNN with seven layers. There are three convolution layers, two subsampling layers,...
Semantic segmentation is an important step of visual scene understanding for autonomous driving. Recently, Convolutional Neural Network (CNN) based methods have successfully applied in semantic segmentation using narrow-angle or even wide-angle pinhole camera. However, in urban traffic environments, autonomous vehicles need wider field of view to perceive surrounding things and stuff, especially at...
Today Unmanned Aerial Systems (UAS) are widely used for many applications that involve advanced payload as is found to be the case for mounted remote sensing apparatus. Remote sensing from UAS platforms is now common and the use of light and smart multi/hyper-spectral cameras has opened the field to novel applications. These sensors can operate in cloudy conditions ensuring ultra high resolution images...
The authors propose a variational level set image segmentation method for intensity inhomogeneous texture image. The method first extracts the main image structure by a relative total variation image decomposition method, which can better decompose the image into structural and textural parts. Then only uses the structural part as the input image for the variational level set segmentation. The intensity...
The work solves the problem of human head detection on the images based on the structural approach. Development of the graphs comparison algorithm by their embedding in the vector subspace is considered. Special points are obtained on the images. These points are clusterized to select characteristic regions of the human's face. Centers of clusters mass are further used as graph points. The graph is...
Image edge information is very important in application areas such as machine learning, image processing, stereo vision, object tracking and pattern recognition. Intensity discontinuities or sudden intensity changes in a region are indicative of the edge region in that region. Although there are many approaches to detecting edge, generally intensity discontinuities or sudden intensity changes in a...
Robust scene understanding of outdoor environments using passive optical sensors is a onerous and essential task for autonomous navigation. The problem is heavily characterized by changing environmental conditions throughout the day and across seasons. Robots should be equipped with models that are impervious to these factors in order to be operable and more importantly to ensure safety in the real-world...
We present a new image feature detection method. Our method selects features based on segmenting points with high local intensity variations across different scales using a robust rank order statistics approach. Our method produces a large number of repeatable features that are invariant to several image transformations such as rotation, scaling, viewpoint, and lighting variations. We show the advantages...
The given work describes a new technique of image segmentation, in particular for building detection on radar or infrared Earth-observation images. The method is based on property of most man-made objects which consist in straight edges and mostly right angles. The developed 2D adaptive image filter assists to detect straight edges even if given image fragment has a low contrast and has been extremely...
Segmenting a structural magnetic resonance imaging (MRI) scan is an important pre-processing step for analytic procedures and subsequent inferences about longitudinal tissue changes. Manual segmentation defines the current gold standard in quality but is prohibitively expensive. Automatic approaches are computationally intensive, incredibly slow at scale, and error prone due to usually involving many...
In this paper we propose a new method for scene representation and recognition based on the concept of Region Subspaces. Each image is pre-segmented into semantically meaningful regions and local features are extracted at different scales from each such region. The Region Subspaces are the low-dimensional linear subspaces calculated from the set of local features inside each region. We also define...
Superpixel based methods have recently shown success in scene segmentation and labeling. In scene labeling, a superpixel algorithm is used first to segment the image into visually consistent small regions; then several feature descriptors are computed and classification is performed for each superpixel. In this paper, Kernel Codebook Encoding (KCB) of superpixel features is proposed. In KCB feature...
K-means is a compute-intensive iterative algorithm, each iteration consists of two steps data assignment and K centroids recalculation. In order to accelerate the compute-intensive portions of k-means, the data assignment and K centroids recalculation steps are offloaded to the GPU in parallel. Only the initialization and convergence tests steps are performed by the CPU. In addition this new version...
Recent progress in Thermal and infrared Non-Destructive Testing (IRNDT) in different fields have provided interesting defect detection solutions. Principal Component Analysis (PCA) based K-means clustering have been successfully introduced and used in many clustering applications. However, PCA suffers from being relatively more sensitive to the noise due to having a linear transformation. On the other...
Image procession algorithms for compensation of scattered radiation influence in X-ray imaging were proposed, studied and optimized by numerical simulations. The algorithms include scattering estimation by convolution (superposition) technique, estimation of kernel functions by Monte-Carlo (MC) simulations, determination the optimal number and shape of kernel functions and images segmentation. Determination...
Co-saliency detection aims at finding the common salient objects in multiple images. In this paper, we introduce a new co-saliency detection model, which includes two main parts: co-salient seed selection using the inter-object recurrence cues from multiple images and saliency label propagation using partially absorbing random walk. With the guidance of co-salient seeds, salient objects are individually...
In this work we address the problem of blind deblurring using a single space-variantly defocused image containing text. We estimate both the all-in-focus image and the blur map corresponding to the space-variant point spread function of the finite aperture camera. Since this problem is highly ill-posed we exploit a recently proposed technique [1] to obtain an initial estimate of the space-variant...
Set the date range to filter the displayed results. You can set a starting date, ending date or both. You can enter the dates manually or choose them from the calendar.