Edge Representation Research Articles

Real-time multi-resolution edge detection with pattern analysis on graphics processing unit

Edge detection is defined as the process of detecting and representing the presences of and locations of image signal discontinuities, which serves as the basic transformation of signals into symbols and it influences the performance of subsequent higher level pattern analysis downstream. In general, the edge detection operation has two main steps: filtering, and detection and localization. In the first step, finding an optimal scale of the filter is an ill-posed problem, especially when a single—global—scale is used over the entire image. Multi-resolution description of the image which can represent the image features occurring in a range of scales is used, where a combination of Gaussian filters with different scales can ameliorate the single scale issue. In the second step, often edge detectors have been designed to capture simple ideal step functions in image data, but real image signal discontinuities deviate from this ideal form. Another three types of deviations from the step function which relate to real distortions occurring in natural images are examined. These types are impulse, ramp, and sigmoid functions which, respectively, represent narrow line signals, simplified blur effects, and more accurate blur modeling. General rules for edge representation based upon the classification of edge types into four categories—ramp, impulse, step, and sigmoid (RISS) are developed from this analysis. Additionally, the proposed algorithm performs connectivity analysis (CA) on edge map to ensure that small, disconnected edges are removed. The performance analysis on experiments supports that the proposed multi-resolution edge detection algorithm with edge pattern analysis does lead to more effective edge detection and localization with improved accuracies. To expand the proposed algorithm into real-time applications, a parallel implementation on a graphics processing unit (GPU) is presented in this paper. For the various configurations in our test, the GPU accelerated RISS shows a scalable speedup as the resolution of an image increases. We also achieved 20 frames per second in video processing (\(640\times 480\)).

Uniform surface-to-line integral reduction of physical optics for curved surfaces by Keller-type equivalent edge currents in MER: Behaviors of fields on reflection shadow boundary

The Modified Edge Representation (MER) is a procedure for defining the equivalent edge currents (EECs) to be used in the surface-to-line integral reduction for computing the physical optics radiation integrals in diffraction analysis. The physical optics surface integration is transformed into the line integration of MER-EECs rigorously for the planar surfaces. It works with the remarkable accuracy even for the curved surface. In contrast to conventional EECs, those in MER are defined without ambiguity not only at the diffraction points but also at arbitrary ones along the periphery. One remarkable advantage is that the MER-EECs, after integrated along periphery, provide uniform fields across the geometrical shadow boundaries though the EECs in the integrand consist of nonuniform Keller-type diffraction coefficients. As the starting point of the applicability check of MER for the curved surfaces, this paper investigates the MER diffraction field behaviors on and near the reflection shadow boundary (RSB). First, the stabilities or the robustness of the numerical line integration are interpreted. Second, the dependence of the RSB field errors upon the curvature of the surfaces is investigated. They become smaller in higher frequency and are related with the aberration of the geometrical reflected ray; for the dipole wave illumination, the error is generally smaller and larger for the concave and convex surfaces, respectively. This surface shape dependence of MER errors in diffraction is quite analogous to those in predicting geometrical optics contributions, the latter of which is the complement of the former and was previously reported by the authors.

Novel Approach of Edges Detection for Digital Images Based On Hybrid Types of Entropy

Edges detection of digital images is used in a various fields of applications ra nging from real-time video surveillance and traffic management to medical imaging applications. Most of the class ical methods for edge detection are based on the first and second order derivatives of gray levels of the pixels of the original ima ge. These processes give rise to the exponential increment of computational time. This paper shows the new algorithm based on both the Tsallis entropy and the Shannon entropy together for edge detection using split and merge technique. The objective is to find the b est edge representation and minimize the computation time. A set of experiments in the domain of edge detection are presented. An edge detection performance compared to the previous classic methods, such as Canny, LOG, and Sobel. Analysis show that the effect of the proposed method is better than those methods in execution time and also is considered as easy implementation

A New CBIR Approach for the Annotation of Medical Images

In the medical field, images, and especially digital images, are produced in ever increasing quantities and used for diagnostics and therapy. Imaging has occupied a huge role in the management of patients, whether hospitalized or not. This gave birth of the annotation of medical image process. The annotation is intended to image analysis and solve the problem of semantic gap. Physicians and radiologists feel better while using annotation techniques for faster making decision and giving solutions to patients in a faster and more accurate way. However, medical images annotation still a hard task specially the process based Content-based image retrieval (CBIR). Recently, advances in Content Based Image Retrieval prompted researchers towards new approaches in information retrieval for image databases. In medical applications it already met some degree of success in constrained problems. For this reason, we focus in this paper on presenting to provide an efficient semi-automatic tool which is used for efficient medical image retrieval from a huge content of medical image database and which is used for further medical diagnosis purposes for the new image annotation, because, efficient content-based image Retrieval in the medical domain is still a challenging problem. The goal of this work is to propose an approach able to compute similarity between a new medical image and old stored images. The annotator has to choose then one of the similar images and annotations related to the selected one are assigned to the new one. The idea is to apply an edge detector algorithm (Sobel algorithm) to the image and extract features from the filtered image by a color histogram. The edge to the image become likes Finger print to a human in our work. It is a search based edge. Edge representation of an image drastically reduces the amount of data to be processed, yet it retains important information about the shapes of objects in the scene. Edges in images constitute an important feature to represent their content and extraction features from filtered image improve searching of similar images, and keeping in the same time the properties of each image. The similarity measurement between images is developed based the Euclidean distance. The method can answer queries by example. The efficiency and performance of the presented method has been evaluated using the precision and the recall. The results of our experiments show high percentage of success, which is satisfactory.

Reconstructing an image from its edge representation

In this paper, we show that a new edge detection scheme developed from the notion of transition in nonlinear physics, associated with the precise computation of its quantitative parameters (most notably singularity exponents) provide enhanced performances in terms of reconstruction of the whole image from its edge representation; moreover it is naturally robust to noise. The study of biological vision in mammals state the fact that major information in an image is encoded in its edges, the idea further supported by neurophysics. The first conclusion that can be drawn from this stated fact is that of being able to reconstruct accurately an image from the compact representation of its edge pixels. The paper focuses on how the idea of edge completion can be assessed quantitatively from the framework of reconstructible systems when evaluated in a microcanonical formulation; and how it redefines the adequation of edge as candidates for compact representation. In the process of doing so, we also propose an algorithm for image reconstruction from its edge feature and show that this new algorithm outperforms the well-known ‘state-of-the-art’ techniques, in terms of compact representation, in majority of the cases.

Network approaches to the genetic dissection of phenotypes in animals and humans

Various genome-wide approaches to identifying the genetic components that underlie phenotypes in animals and humans have been developed during the last several decades. The relationship between gene and phenotype, however, cannot be represented by a simple one-to-one correspondence. Rather, many genes are typically related to a single phenotype and many phenotypes can be associated with a single gene, a major theme within the study of complex phenotypes. Therefore, to dissect the genetics of complex phenotypes, one must not only identify the genetic components involved but also the relationships between genes. To fulfill this new goal in modern genetics, the field of network science has recently tackled the complexity of phenotypes. There are various types of gene networks, which are defined by their differential representation of network edges (i.e., relationships). Different networks map physical, genetic, functional, and regulatory interactions between genes. Gene networks can be constructed using a wide variety of experimental and computational methods, which provide complimentary information about the genetic organization of phenotypes. The predictive power of a gene network is further augmented via integration with functional genomics or genetics data, including expression, loss-of-function, or chromosomal interval or nucleotide position data associated with a phenotype. Although the field of network-based genetics has made phenomenal progress during the last decade, many limitations, such as the completeness and dynamicity of gene networks, must still be overcome.

Enhanced representation of spectral contrasts in the primary auditory cortex

The role of early auditory processing may be to extract some elementary features from an acoustic mixture in order to organize the auditory scene. To accomplish this task, the central auditory system may rely on the fact that sensory objects are often composed of spectral edges, i.e., regions where the stimulus energy changes abruptly over frequency. The processing of acoustic stimuli may benefit from a mechanism enhancing the internal representation of spectral edges. While the visual system is thought to rely heavily on this mechanism (enhancing spatial edges), it is still unclear whether a related process plays a significant role in audition. We investigated the cortical representation of spectral edges, using acoustic stimuli composed of multi-tone pips whose time-averaged spectral envelope contained suppressed or enhanced regions. Importantly, the stimuli were designed such that neural responses properties could be assessed as a function of stimulus frequency during stimulus presentation. Our results suggest that the representation of acoustic spectral edges is enhanced in the auditory cortex, and that this enhancement is sensitive to the characteristics of the spectral contrast profile, such as depth, sharpness and width. Spectral edges are maximally enhanced for sharp contrast and large depth. Cortical activity was also suppressed at frequencies within the suppressed region. To note, the suppression of firing was larger at frequencies nearby the lower edge of the suppressed region than at the upper edge. Overall, the present study gives critical insights into the processing of spectral contrasts in the auditory system.

Cognitive physics-based method for image edge representation and extraction with uncertainty

Image edge detection is an important tool of image processing, in which edge representation and extraction with uncertainty is one of key issues. Based on the physics-like methods for image edge representation and extraction, a novel cognitive physics-based method with uncertainty is proposed. The method uses data field to discover the global information from the image and then to map it from grayscale space to the appropriate potential space. From the point of view of the field theory, the method establishes an extensible theoretical framework and unifies the existing physics-like methods. On the other hand, the method defines the ascending half-cloud to construct the internal relationship between the range of cloud uncertainty degree and the edge representation and extraction. Finally, the method achieves image edge representation and extraction with uncertainty using the cognitive physics. The time complexity of the proposed algorithm is approximately linear in the size of the original image. It is indicated by the quantitative and qualitative experiments that the proposed method yields accurate and robust result, and is reasonable and effective.

Curvelet image denoising of mammogram images

Mammography, the most commonly used diagnostic technique is used for early detection of breast cancer. As mammograms are low contrast and noisy images, it is essential to reduce noise while preserving fine details and edges. In order to obtain efficient diagnosis, a constructive analysis curvelet is used to provide optimal sparse representation of smooth objects and edges. Hence, in this work, a novel noise reduction method using curvelet transform is proposed for digital mammogram images. Here, digital curvelet transform is applied to the mammogram images which are embedded in random, salt and pepper, Poisson, speckle and Gaussian noises. As curvelet coefficients occur at all scales, locations and orientations, proposed thresholding algorithm applied at the expansions lead to better denoising. And it also offers exact reconstruction of images that exhibit higher perceptual quality in recovery of edges. The PSNR values of the enhanced images are higher than related earlier techniques.

Geomorphic coupling and sediment connectivity in an alpine catchment — Exploring sediment cascades using graph theory

Through their relevance for sediment budgets and the sensitivity of geomorphic systems, geomorphic coupling and (sediment) connectivity represent important topics in geomorphology. Since the introduction of the systems perspective to physical geography by Chorley and Kennedy (1971), a catchment has been perceived as consisting of landscape elements (e.g. landforms, subcatchments) that are coupled by geomorphic processes through sediment transport. In this study, we present a novel application of mathematical graph theory to explore the network structure of coarse sediment pathways in a central alpine catchment. Numerical simulation models for rockfall, debris flows, and (hillslope and channel) fluvial processes are used to establish a spatially explicit graph model of sediment sources, pathways and sinks. The raster cells of a digital elevation model form the nodes of this graph, and simulated sediment trajectories represent the corresponding edges. Model results are validated by visual comparison with the field situation and aerial photos. The interaction of sediment pathways, i.e. where the deposits of a geomorphic process form the sources of another process, forms sediment cascades, represented by paths (a succession of edges) in the graph model. We show how this graph can be used to explore upslope (contributing area) and downslope (source to sink) functional connectivity by analysing its nodes, edges and paths. The analysis of the spatial distribution, composition and frequency of sediment cascades yields information on the relative importance of geomorphic processes and their interaction (however regardless of their transport capacity). In the study area, the analysis stresses the importance of mass movements and their interaction, e.g. the linkage of large rockfall source areas to debris flows that potentially enter the channel network. Moreover, it is shown that only a small percentage of the study area is coupled to the channel network which itself is longitudinally disconnected by natural and anthropogenic barriers. Besides the case study, we discuss the methodological framework and alternatives for node and edge representations of graph models in geomorphology. We conclude that graph theory provides an excellent methodological framework for the analysis of geomorphic systems, especially for the exploration of quantitative approaches towards sediment connectivity.

Image-radargram analysis based on generalized Hough transform: experimental cases

This paper presents an algorithm based on the generalized Hough transform (GHT) to enhance geological structures in a 2D profile image constructed from ground-penetrating radar data. The proposed algorithm is applied over an edge representation image of the radargram; its local application (in Windows) allows us to identify structures and to obtain a global image delimiting the subsurface structures of interest. To show and validate the program, it was applied to two experimental cases: the first one is a profile of volcanic stratigraphic structures, and the second one consists of three buried pipes. The GHT algorithm enhances the main geological structures for both examples. Due to the use of a monostatic antenna in the case of buried pipes, a proper velocity analysis could not be done. As a consequence, the radargram remained under-migrated leaving some traces of diffraction hyperbola tails. These traces were detected by the GHT; therefore, a robust algorithm was designed to adjust a circle over these hyperbola traces. This secondary algorithm is called ‘circular structure determination’.

Score-Level Fusion of Local Spatial, Scale and Directional Features Based Face Recognition Approach for Single Sample Problem

The local appearance based methods have been successfully applied to face recognition and achieved state-of-the-art performance. In this paper we propose a face representation approach which explores the information in the spatial domain along with score-level fusion of directional features in different scales. The local spatial texture features are extracted using the local MinMax binary pattern (LMinMaxBP) approach as well as with the help of the selective local texture feature approach. The dual tree complex wavelet transform (DT-CWT) provides a local multiscale description of images with good directional selectivity, effective edge representation and invariance to shifts and in-plane rotations. It is insensitive to illumination variations and facial expression changes. 2-D dual tree complex wavelet transform is less redundant and computationally efficient. The fusion of local DT-CWT coefficients of detail subbands are used to extract the multi scale facial features with good directional selectivity which improves the face recognition with small sample size in less computation. This score-level fusion approach combines information from different domains to give a good face representation for recognition. Extensive experimental results on FERET, ORL, Yale and Indian Face databases show the significant advantages of the proposed method over the existing ones.

Misalignment-robust, edge-based image fusion method

We propose an image fusion method robust to misaligned source images based on their multiscale edge representations. Significant long edge curves at the second scale are selected to decide edge locations at each scale for the multiscale edge representations of source images. Then, processes are only executed on the representations that contain the main spatial structures of the images and also help suppress noise interference. A registration process is embedded in our fusion method. Edge correlation, calculated at the second scale, is involved as a match measure determining the fusion rules and also as a similarity measure quantifying the matching extent between source images, which makes the registration and fusion processes share the same data and hence lessens the computation of our method. Experimental results prove that, no matter whether in a noiseless or noisy condition, the proposed method provides satisfying treatment to misregistered source images and behaves well in terms of visual and objective evaluations on the fusion results, which further verifies the robustness of our edge-based method to misregistration and noise.

Difference of scattering geometrical optics components and line integrals of currents in modified edge representation

Equivalent edge currents (EECs) are widely used to asymptotically extract and express the diffraction from the periphery of the scatterers, as in the concept suggested by Young. Authors proposed novel EECs based on the unique concept named as Modified Edge Representation (MER), for surface to line integral reduction of Physical Optics (PO) radiation integral. MER is based upon not only asymptotic approximation but also Stokes' theorem, and it has remarkable accuracy even for small scatterers and is uniformly applicable at the geometrical boundaries. Moreover MER‐EECs are clearly defined not only at the edges but also everywhere on the scatterer surface, with the singularity at the stationary phase point (SPP) if any. It comes that the radiation integral in general, consisting of both the diffraction and the Geometrical Optics (GO) components, could be reduced into two sets of line integration of MER‐EECs along the periphery and the indentation at SPPs. In this paper, the GO reflection is numerically compared with the MER indentation integral around the reflected point for the dipole scattering from an ellipsoid; the error is empirically derived in analytical form.

VEHICLE TYPE AND MAKE RECOGNITION BY COMBINED FEATURES AND ROTATION FOREST ENSEMBLE

Vehicle type/make recognition based on images captured by surveillance cameras is a challenging task in intelligent transportation system and automatic surveillance. In this paper, we comparatively studied two feature extraction methods for image description, i.e. a new multiresolution analysis tool called Fast Discrete Curvelet Transform and the pyramid histogram of oriented gradients (PHOG). Curvelet Transform has better directional and edge representation abilities than widely used wavelet transform, which is particularly appropriate for the description of images rich with edges. PHOG represents the local shape by a histogram of edge orientations computed for each image sub-region, quantized into a number of bins, thus has the ascendency in its description of more discriminating information. A composite feature description from PHOG and Curvelet can further increase the accuracy of classification by taking their complementary information. We also investigated the applicability of the Rotation Forest (RF) ensemble method for vehicle classification based on the combined features. The RF ensemble contains a set of base multilayer perceptrons which are trained using principal component analysis to rotate the original axes of combined features of vehicle images. The class label is assigned by the ensemble via majority voting. Experimental results using more than 600 images from 21 makes of cars/vans show the effectiveness of the proposed approach. The composite feature is better than any single feature in the classification accuracy and the ensemble model produces better performance compared to any of the individual neural network base classifier. With a moderate ensemble size of 20, the Rotation Forest ensembles offers a classification rate close to 96.5%, exhibiting promising potentials for real-life applications.

Infrared Image And Visible Light Image Fusion Based On Nonsubsampled Contourlet Transform And The Gradient Of Uniformity

The nonsubsampled contourlet transform (NSCT) is a new multi-resolution transform, which can give an asymptotic representation of edges and contours in image by virtue of its characteristics of multi-direction, flexible multi-scale and shift-invariant simultaneity. For fusion between visible light image and infrared image, this paper proposed a new method based on nonsubsampled contourlet transform and the visual characteristics of region. Firstly, two original images were decomposed into different frequency sub-band coefficients by using NSCT, and then the selection of the low frequency subband coefficient and the bandpass direction subband coefficient was discussed respectively. For the choice of low frequency subband coefficient, this paper proposed a new method that chose coefficient based on the visual characteristics of region and gradient characteristics. For the choice of bandpass direction subband coefficient, this paper uses a direct method that chose the coefficient based on maximum regional uniformity obtained by the visual characteristics of region. Finally, the fused image was obtained by performing the inverse NSCT on the combined coefficients. The experimental results show that this fusion algorithm proposed in this paper achieved significant results to extract target in the infrared image and preserve edge of the target. At the same time, it also achieved good results to effectively preserve image details in the visible light image.

Feature Image Generation Using Low, Mid and High Frequency Regions for Face Recognition

In this paper, we propose a generation of feature image for face recognition. Feature image is generated using Discrete Cosine Transform (DCT), which tries to best classify different face images by maximizing the individual difference between face images. Low, Mid, and High frequency components of DCT are used to hold different information of face by providing a local description of different facial components such as detailed smooth region description and, effective edge representation. Reduction in different frequency components is achieved using statistical measure such as standard deviation. A comparison of proposed work with other standard methods is done on ORL face datasets. An experimental result shows a significant improvement of the proposed scheme.

Image Fusion Method based on Non-Subsampled Contourlet Transform

Considering human visual system and characteristics of images, a novel image fusion strategy is presented for panchromatic high resolution image and multispectral image in non-subsampled contourlet transform (NSCT) domain. The NSCT can give an asymptotic optimal representation of edges and contours in image by virtue of its characteristics of good multiresolution, shiftinvariance, and high directionality. An intensity component addition strategy based on NMF algorithm is introduced into NSCT domain to preserve spatial resolution and color content. Experiments show that the proposed algorithm can not only reduce computational complexities, but achieve better performances than other mentioned techniques both in visual point and statistics compared with the traditional principle component analysis (PCA) method, intensity-hue-saturation (IHS) transform technique, wavelet transform weighted fusion method, corresponding wavelet transform-based fusion method, and contourlet transform-based fusion method.

Directional lifting wavelet and universal trellis-coded quantisation-based image coding algorithm and objective quality evaluation

In this study, an image coding algorithm based on directional lifting wavelet transform (DLWT) and universal trellis-coded quantisation (UTCQ) is presented, and the coding performance is evaluated with three objective image quality metrics. Compared to the discrete wavelet transform, DLWT performing prediction and update along the direction of the local region can provide an efficient representation of edges in images, but shows a similar ability in representing the smooth region. To further improve the visual quality of the smooth background regions, UTCQ is adopted to quantising the wavelet coefficients. The proposed algorithm is measured with not only the dominant peak signal-to-noise ratio (PSNR), but also new metrics multi-scale structural similarity index measure (MSSIM) and visual information fidelity (VIF) which provide a better approximation to the perceived image quality than PSNR by taking the property of human visual system (HVS) into account. Experimental results show that the proposed algorithm has the best MSSIM and VIF performance among the compared algorithms (including JPEG2000) for the typical test images, and its decoded images at low bit-rate are visually more appealing in both edges and smooth background regions. For image Barbara, the proposed algorithm outperforms JPEG2000 up to 24.63% relatively in VIF and 1.93%dB in PSNR at 0.5%bpp, at most 3.62% relatively in MSSIM at 0.125%bpp. The experimental results also show that UTCQ does perform better than scalar quantisation (SQ) in MSSIM and VIF and improves the subjective visual quality, although UTCQ is not necessarily better than SQ in PSNR.

Treevis.net: a tree visualization reference.

Tree visualization is one of the best-studied areas of information visualization; researchers have developed more than 200 visualization and layout techniques for trees. The treevis.net project aims to provide a hand-curated bibliographical reference to this ever-growing wealth of techniques. It offers a visual overview that users can filter to a desired subset along the design criteria of dimensionality, edge representation, and node alignment. Details, including links to the original publications, can be brought up on demand. Treevis.net has become a community effort, with researchers sending in preprints of their tree visualization techniques to be published or pointing out additional information.