scispace - formally typeset
Search or ask a question

Showing papers on "Distance transform published in 2014"


Proceedings ArticleDOI
10 Nov 2014
TL;DR: A novel image representation method by learning and using kernel classifiers using the one-against-all rule and the Euclidean distance between the classification response vectors is used as the new similarity measure.
Abstract: The learning of image representation is always the most important problem in computer vision community. In this paper, we propose a novel image representation method by learning and using kernel classifiers. We firstly train classifiers using the one-against-all rule, then use them classify the candidate images, and finally using the classification responses as the new representations. The Euclidean distance between the classification response vectors are used as the new similarity measure. The experimental results from a large scale image database show that the proposed algorithm can outperform the original feature on image retrieval problem.

230 citations


Proceedings ArticleDOI
23 Jun 2014
TL;DR: This work proposes a robust and accurate method to extract the centerlines and scale of tubular structures in 2D images and 3D volumes by reformulating centerline detection in terms of a regression problem.
Abstract: We propose a robust and accurate method to extract the centerlines and scale of tubular structures in 2D images and 3D volumes. Existing techniques rely either on filters designed to respond to ideal cylindrical structures, which lose accuracy when the linear structures become very irregular, or on classification, which is inaccurate because locations on centerlines and locations immediately next to them are extremely difficult to distinguish. We solve this problem by reformulating centerline detection in terms of a regression problem. We first train regressors to return the distances to the closest centerline in scale-space, and we apply them to the input images or volumes. The centerlines and the corresponding scale then correspond to the regressors local maxima, which can be easily identified. We show that our method outperforms state-of-the-art techniques for various 2D and 3D datasets.

128 citations


Patent
21 Feb 2014
TL;DR: In this paper, a system for generating compressed light field representation data using captured light fields in accordance with the embodiment of the invention is described. But the system is limited to a set of images including a reference image and at least one alternate view image.
Abstract: Systems and methods for the generating compressed light field representation data using captured light fields in accordance embodiments of the invention are disclosed. In one embodiment, an array camera includes a processor and a memory connected configured to store an image processing application, wherein the image processing application configures the processor to obtain image data, wherein the image data includes a set of images including a reference image and at least one alternate view image, generate a depth map based on the image data, determine at least one prediction image based on the reference image and the depth map, compute prediction error data based on the at least one prediction image and the at least one alternate view image, and generate compressed light field representation data based on the reference image, the prediction error data, and the depth map.

83 citations


Journal ArticleDOI
TL;DR: A language independent global method for automatic text line extraction that computes an energy map of a text image and determines the seams that pass across and between text lines, and develops two algorithms along this novel idea.

80 citations


Book ChapterDOI
07 May 2014
TL;DR: A robust and efficient automatic approach to define and compute a signed distance field for arbitrary triangular geometry and proves that exterior grid points can reuse a shifted original unsigned distance field, whereas for interior cells, the signed field is computed from the offset surface geometry.
Abstract: Many meshes in computer animation practice are meant to approximate solid objects, but the provided triangular geometry is often unoriented, non-manifold or contains self-intersections, causing inside/outside of objects to be mathematically ill-defined. We describe a robust and efficient automatic approach to define and compute a signed distance field for arbitrary triangular geometry. Starting with arbitrary (non-manifold) triangular geometry, we first define and extract an offset manifold surface using an unsigned distance field. We then automatically remove any interior surface components. Finally, we exploit the manifoldness of the offset surface to quickly detect interior distance field grid points. We prove that exterior grid points can reuse a shifted original unsigned distance field, whereas for interior cells, we compute the signed field from the offset surface geometry. We demonstrate improved performance both using exact distance fields computed using an octree, and approximate distance fields computed using fast marching. We analyze the time and memory costs for complex meshes that include self-intersections and non-manifold geometry. We demonstrate the effectiveness of our algorithm by using the signed distance field for collision detection and generation of tetrahedral meshes for physically based simulation.

49 citations


Journal ArticleDOI
TL;DR: In this article, a sparse representation-based approach is proposed to encode the information content of an image using information from the other image, and uses the compactness (sparsity) of the representation as a measure of its compressibility.
Abstract: A new line of research uses compression methods to measure the similarity between signals. Two signals are considered similar if one can be compressed significantly when the information of the other is known. The existing compression-based similarity methods, although successful in the discrete one dimensional domain, do not work well in the context of images. This paper proposes a sparse representation-based approach to encode the information content of an image using information from the other image, and uses the compactness (sparsity) of the representation as a measure of its compressibility (how much can the image be compressed) with respect to the other image. The sparser the representation of an image, the better it can be compressed and the more it is similar to the other image. The efficacy of the proposed measure is demonstrated through the high accuracies achieved in image clustering, retrieval and classification.

49 citations


Journal ArticleDOI
TL;DR: It is shown that the considered distances have a number of appealing theoretical properties and exhibit very good performance in template matching and object classification for fuzzy segmented images as well as when applied directly on gray-level intensity images.
Abstract: We present four novel point-to-set distances defined for fuzzy or gray-level image data, two based on integration over α-cuts and two based on the fuzzy distance transform. We explore their theoretical properties. Inserting the proposed point-to-set distances in existing definitions of set-to-set distances, among which are the Hausdorff distance and the sum of minimal distances, we define a number of distances between fuzzy sets. These set distances are directly applicable for comparing gray-level images or fuzzy segmented objects, but also for detecting patterns and matching parts of images. The distance measures integrate shape and intensity/membership of observed entities, providing a highly applicable tool for image processing and analysis. Performance evaluation of derived set distances in real image processing tasks is conducted and presented. It is shown that the considered distances have a number of appealing theoretical properties and exhibit very good performance in template matching and object classification for fuzzy segmented images as well as when applied directly on gray-level intensity images. Examples include recognition of hand written digits and identification of virus particles. The proposed set distances perform excellently on the MNIST digit classification task, achieving the best reported error rate for classification using only rigid body transformations and a kNN classifier.

44 citations


Journal ArticleDOI
TL;DR: This paper presents a polynomial time algorithm, that provably calculates the exact values of MBD for the digital images and compares this new algorithm, theoretically and experimentally, with the algorithm presented in [1] , which computes the approximate values of the MBD.

42 citations


Patent
21 Feb 2014
TL;DR: In this article, a system for generating compressed light field representation data using captured light fields in accordance with the embodiment of the invention is described. But the system is limited to a set of images including a reference image and at least one alternate view image.
Abstract: Systems and methods for the generating compressed light field representation data using captured light fields in accordance embodiments of the invention are disclosed. In one embodiment, an array camera includes a processor and a memory connected configured to store an image processing application, wherein the image processing application configures the processor to obtain image data, wherein the image data includes a set of images including a reference image and at least one alternate view image, generate a depth map based on the image data, determine at least one prediction image based on the reference image and the depth map, compute prediction error data based on the at least one prediction image and the at least one alternate view image, and generate compressed light field representation data based on the reference image, the prediction error data, and the depth map.

36 citations


Proceedings ArticleDOI
06 Jul 2014
TL;DR: The experimental result shows the proposed method of plant recognition based on leaf image gets the better accuracy of recognition than other methods.
Abstract: Plant recognition recently becomes more and more attractive in computer vision and pattern recognition. Although some researchers have proposed several methods, their accuracy is not satisfactory. Therefore, a novel method of plant recognition based on leaf image is proposed in the paper. Both shape and texture features are employed in the proposed method Texture feature is extracted by intersecting cortical model, and shape feature is obtained by the representation of center distance sequence. Support vector machine is employed for the classifier. The leaf image is preprocessed to get better quality for extracting features, and then entropy sequence and center distance sequence are obtained by intersecting cortical model and center distance transform, respectively. Redundant data of entropy sequence vector and center distance are reduced by principal component analysis. Finally, feature vector is imported into the classifier for classification. In order to evaluate the performance, several existing methods are used to compare with the proposed method and three leaf image datasets are taken as test samples. The experimental result shows the proposed method gets the better accuracy of recognition than other methods.

33 citations


Journal ArticleDOI
TL;DR: A new focus detection based multi-focus image fusion algorithm is proposed in which pixels in smooth parts are treated differently according to their classification, and the fused image is achieved with the assistance of fusing map.

Proceedings ArticleDOI
23 Jun 2014
TL;DR: This paper uses the well known Riemannian framework never before used for point cloud matching, and presents a novel matching algorithm that outperforms state-of-the-art point set registration algorithms on many quantitative metrics.
Abstract: In this paper, we cast the problem of point cloud matching as a shape matching problem by transforming each of the given point clouds into a shape representation called the Schrodinger distance transform (SDT) representation. This is achieved by solving a static Schrodinger equation instead of the corresponding static Hamilton-Jacobi equation in this setting. The SDT representation is an analytic expression and following the theoretical physics literature, can be normalized to have unit 2 norm -- making it a square-root density, which is identified with a point on a unit Hilbert sphere, whose intrinsic geometry is fully known. The Fisher-Rao metric, a natural metric for the space of densities leads to analytic expressions for the geodesic distance between points on this sphere. In this paper, we use the well known Riemannian framework never before used for point cloud matching, and present a novel matching algorithm. We pose point set matching under rigid and non-rigid transformations in this framework and solve for the transformations using standard nonlinear optimization techniques. Finally, to evaluate the performance of our algorithm -- dubbed SDTM -- we present several synthetic and real data examples along with extensive comparisons to state-of-the-art techniques. The experiments show that our algorithm outperforms state-of the-art point set registration algorithms on many quantitative metrics.

Proceedings ArticleDOI
20 May 2014
TL;DR: The proposed method is shown to be invariant to image transformations (translation, rotation, reflection and scaling) and robust to minor deformations and occlusions and is used as a classifier for plant leaf classification.
Abstract: In this paper, we use centroid distance and axis of least inertia method for plant leaf classification. For this propose the RGB (Red, Green, Blue) image are converted to the binary image. Then, Canny operator is applied to the binary image to recognize the edges of the image before thinning the edges. After that, the boundary of the image is traced to sample the shape. Sampling helps us to avoid time-consuming computations. We compute the centroid distance of these points and distance of sampling points from axis of least inertia line. By selecting a fixed start point and normalizing the distances, the proposed method is shown to be invariant to image transformations (translation, rotation, reflection and scaling) and robust to minor deformations and occlusions. In this study, probabilistic neural network (PNN) has been used as a classifier. Two public leaf datasets including: Swedish leaf dataset and Flavia dataset are evaluated. Experimental results demonstrate the superior performance of the proposed feature in plant leaf classification.

Journal ArticleDOI
TL;DR: In this paper, the authors presented different improvements of ODSIM methodology for simulating more realistic shapes in the particular case of karst, using a custom distance field computed with a fast marching method.

Journal ArticleDOI
TL;DR: This work uses a GPU to build an exact adaptive distance field, constructed from an octree by using the Morton code, and uses rectangle-swept spheres to construct a bounding volume hierarchy (BVH) around a triangulated model.
Abstract: Most techniques for real-time construction of a signed distance field, whether on a CPU or GPU, involve approximate distances. We use a GPU to build an exact adaptive distance field, constructed from an octree by using the Morton code. We use rectangle-swept spheres to construct a bounding volume hierarchy (BVH) around a triangulated model. To speed up BVH construction, we can use a multi-BVH structure to improve the workload balance between GPU processors. An upper bound on distance to the model provided by the octree itself allows us to reduce the number of BVHs involved in determining the distances from the centers of octree nodes at successively lower levels, prior to an exact distance query involving the remaining BVHs. Distance fields can be constructed 35-64 times as fast as a serial CPU implementation of a similar algorithm, allowing us to simulate a piece of fabric interacting with the Stanford Bunny at 20 frames per second.

Book ChapterDOI
06 Sep 2014
TL;DR: This work develops an uncertainty based representation of line segments in the ground image and incorporates it into a geometric matching framework and shows that this approach is able to rule out a considerable portion of false candidate regions even in a database composed of geographic areas with similar visual appearances.
Abstract: Image based geolocation aims to answer the question: where was this ground photograph taken? We present an approach to geolocalating a single image based on matching human delineated line segments in the ground image to automatically detected line segments in ortho images. Our approach is based on distance transform matching. By observing that the uncertainty of line segments is non-linearly amplified by projective transformations, we develop an uncertainty based representation and incorporate it into a geometric matching framework. We show that our approach is able to rule out a considerable portion of false candidate regions even in a database composed of geographic areas with similar visual appearances.

Journal ArticleDOI
TL;DR: It is shown that FEED class algorithms unite properties of ordered propagation, raster scanning, and independent scanning DT and outperform any other approximate and exact Euclidean DT with its time complexity O(N), even after their optimization.
Abstract: A new unique class of foldable distance transforms of digital images (DT) is introduced, baptized: Fast Exact Euclidean Distance (FEED) transforms. FEED class algorithms calculate the DT starting directly from the definition or rather its inverse. The principle of FEED class algorithms is introduced, followed by strategies for their efficient implementation. It is shown that FEED class algorithms unite properties of ordered propagation, raster scanning, and independent scanning DT. Moreover, FEED class algorithms shown to have a unique property: they can be tailored to the images under investigation. Benchmarks are conducted on both the Fabbri et al. data set and on a newly developed data set. Three baseline, three approximate, and three state-of-the-art DT algorithms were included, in addition to two implementations of FEED class algorithms. It illustrates that FEED class algorithms i) provide truly exact Euclidean DT; ii) do no suffer from disconnected Voronoi tiles, which is a unique feature for non-parallel but fast DT; iii) outperform any other approximate and exact Euclidean DT with its time complexity O(N), even after their optimization; and iv) are unequaled in that they can be adapted to the characteristics of the image class at hand. The source code of all algorithms included as well as the data sets used for both benchmarks are provided as supplementary material to this article.

Book ChapterDOI
28 May 2014
TL;DR: Three types of neighborhood relations are used on the grid, and therefore three weights are used to define a distance function, which is ready for application in image processing and other fields.
Abstract: In this paper we introduce weighted distances on a triangular grid. Three types of neighborhood relations are used on the grid, and therefore three weights are used to define a distance function. Some properties of the weighted distances, including metrical properties are discussed. We also give algorithms that compute the weighted distance of any point-pair on a triangular grid. Formulae for computing the distance are also given. Therefore the introduced new distance functions are ready for application in image processing and other fields.

Journal ArticleDOI
TL;DR: A region-based image fusion algorithm based on the fact that multi-focus images contain compensatory focused regions that can be selected to create a merged image that facilitates accurate detection of fiber edges in a nonwoven structure is presented.
Abstract: The multifocal phenomenon is a common problem when viewing a thick nonwoven sample under a light microscope. Multi-focus image fusion is a technique used to combine a series of partially focused images of the same scene into one fully focused image, and permits the accurate measurement of object features within the scene. This paper presents a region-based image fusion algorithm based on the fact that multi-focus images contain compensatory focused regions that can be selected to create a merged image. The process starts with the selection of a few reliable points with the highest local sharpness values and where there is coherent edge information (object features). Regions are then formed through diffusion or expansion of these selected source points. The final coupled boundaries among the diffusing sources are determined using the distance transform. Once the new image is divided into a number of regions, each region is filled with the corresponding region selected from one of the multi-focus images tha...

Book ChapterDOI
14 Sep 2014
TL;DR: To automatically segment the prostate, the proposed distance transform is more effective than the traditional classification-based method for driving the deformable segmentation, and can achieve more consistent segmentations than human raters, and more accurate results than the existing methods under comparison.
Abstract: Segmenting the prostate from CT images is a critical step in the radiotherapy planning for prostate cancer. The segmentation accuracy could largely affect the efficacy of radiation treatment. However, due to the touching boundaries with the bladder and the rectum, the prostate boundary is often ambiguous and hard to recognize, which leads to inconsistent manual delineations across different clinicians. In this paper, we propose a learning-based approach for boundary detection and deformable segmentation of the prostate. Our proposed method aims to learn a boundary distance transform, which maps an intensity image into a boundary distance map. To enforce the spatial consistency on the learned distance transform, we combine our approach with the auto-context model for iteratively refining the estimated distance map. After the refinement, the prostate boundaries can be readily detected by finding the valley in the distance map. In addition, the estimated distance map can also be used as a new external force for guiding the deformable segmentation. Specifically, to automatically segment the prostate, we integrate the estimated boundary distance map into a level set formulation. Experimental results on 73 CT planning images show that the proposed distance transform is more effective than the traditional classification-based method for driving the deformable segmentation. Also, our method can achieve more consistent segmentations than human raters, and more accurate results than the existing methods under comparison.

Journal ArticleDOI
TL;DR: A purely algebraic, and computationally efficient technique is described for constructing distance measures from Non-Uniform Rational B-Splines boundaries that retain the geometric exactness of the boundaries while eliminating the need for iterative and non-robust distance calculation.

Proceedings ArticleDOI
11 Mar 2014
TL;DR: In this paper, an adaptation of the Dark Channel Prior is proposed to obtain a rough distance map estimative, and some model simplifications are made to obtain the restoration of the image.
Abstract: The underwater vision is highly spoiled by the underwater degradation effects. As light propagates in the water or other participative mediums, it suffers from a substantial scattering effect that produces poor image quality. Based on a physical model that describes this phenomenon it is possible to recover an haze-free image. But, for this procedure to succeed, it is necessary to obtain certain parameters from the model. With an adaptation of the Dark Channel Prior, proposed by this paper, we are able to obtain a rough distance map estimative. With this, and some model simplifications, we are able to successfully obtain the restoration of the image.

Journal ArticleDOI
Dong-Jin Yoo1
TL;DR: Wang et al. as discussed by the authors proposed a new projection image generation algorithm that automatically and robustly generates 2D projection image data using the volumetric distance field and triply periodic minimal surface (TPMS) pore morphology.
Abstract: Advanced additive manufacture (AM) techniques have been developed to generate three-dimensional (3D) tissue scaffolds with complex topography and controlled internal pore architecture. Among the various AM methods, projection stereolithography (PSL) can be used to fabricate intricate 3D tissue scaffolds that can be engineered to mimic the microarchitecture of tissues. PSL system offers the advantages of enhanced fabrication speed and accuracy compared with conventional stereolithography system. To design and fabricate a complex 3D scaffold, we propose a new projection image generation algorithm that automatically and robustly generates 2D projection image data. The method uses the volumetric distance field (VDF) and triply periodic minimal surface (TPMS) pore morphology. By the creative combination of VDF and TPMS-based pore architecture, we can easily and rapidly generate projection image data for PSL system without using complicated 3D scaffold models. An effective Boolean operation based on VDF was utilized to improve the efficiency of geometrical manipulations required in the scaffold projection image generation. The design results demonstrated that the proposed algorithm can completely alleviate all the limitations and problems of the previous approaches mostly based on time-consuming and error-prone slicing process.

DOI
01 Jan 2014
TL;DR: An algorithm for fast continuous collision detection between points and signed distance fields is presented, and how to robustly use it for 6-DoF haptic rendering of contact between objects with complex geometry is demonstrated.
Abstract: We present an algorithm for fast continuous collision detection between points and signed distance fields. Such robust queries are often needed in computer animation, haptics and virtual reality applications, but have so far only been investigated for polygon (triangular) geometry representations. We demonstrate how to use an octree subdivision of the distance field for fast traversal of distance field cells. We also give a method to combine octree subdivision with points organized into a tree hierarchy, for efficient culling of continuous collision detection tests. We apply our method to multibody rigid simulations, and demonstrate that our method accelerates continuous collision detection between points and distance fields by an order of magnitude.

Patent
18 Mar 2014
TL;DR: In this paper, a method and system for occlusion region detection and measurement between a pair of images is described, where a processing device receives a first image and a second image.
Abstract: A method and system for occlusion region detection and measurement between a pair of images are disclosed. A processing device receives a first image and a second image. The processing device estimates a field of motion vectors between the first image and the second image. The processing device motion compensates the first image toward the second image to obtain a motion-compensated image. The processing device compares a plurality of pixel values of the motion-compensated image to a plurality of pixels of the first image to estimate an error field. The processing device inputs the error field to a weighted error cost function to obtain an initial occlusion map. The processing device regularizes the initial occlusion map to obtain a regularized occlusion map.

Journal ArticleDOI
TL;DR: Experimental results show that the proposed stereo matching algorithm based on distance transform to generate high-quality disparity maps with occlusion handling outperforms conventional stereo matching algorithms with occLusion handling.

Patent
30 Jun 2014
TL;DR: In this paper, a computing device performs matching between a target image and one or more template images, where the computing device receives image data and performs an edge detection algorithm on the image data.
Abstract: A computing device performs matching between a target image and one or more template images. The computing device receives image data and performs an edge detection algorithm on the image data. The edge detection algorithm includes a distance metric based on angles between gradient vectors in the image data and gradient vectors in one or more templates. The computing device matches a building model to the image data based on results of the edge detection algorithm, wherein the building model is associated with the one or more templates.

Proceedings ArticleDOI
24 Aug 2014
TL;DR: A new robust approach of extracting arc skeletons for three-dimensional (3-D) elongated fuzzy objects, which avoids spurious branches without requiring post-pruning, is presented.
Abstract: Traditional arc skeletonization algorithms using the principle of Blum's transform, often, produce unwanted spurious branches due to boundary irregularities and digital effects on objects and other artifacts. This paper presents a new robust approach of extracting arc skeletons for three-dimensional (3-D) elongated fuzzy objects, which avoids spurious branches without requiring post-pruning. Starting from a root voxel, the method iteratively expands the skeleton by adding a new branch in each iteration that connects the farthest voxel to the current skeleton using a minimum-cost geodesic path. The path-cost function is formulated using a novel measure of local significance factor defined by fuzzy distance transform field, which forces the path to stick to the centerline of the object. The algorithm terminates when dilated skeletal branches fill the entire object volume or the current farthest voxel fails to generate a meaningful branch. Accuracy of the algorithm has been evaluated using computer-generated blurred and noisy phantoms with known skeletons. Performance of the method in terms of false and missing skeletal branches, as defined by human expert, has been examined using in vivo CT imaging of human intrathoracic airways. Experimental results from both experiments have established the superiority of the new method as compared to a widely used conventional method in terms of accuracy of medialness as well as robustness of true and false skeletal branches.

Journal ArticleDOI
TL;DR: Compared to existing methods, the proposed algorithm calculates geographical distances based on an earth ellipsoid and allows Voronoi generators to take complex shapes and its approximation error is bounded thus enabling users to control the accuracy of the Vor onoi diagram through grid resolution.

Journal ArticleDOI
TL;DR: A novel colorization scheme that takes advantage of the modified morphological distance transform to propagate the color, scribbled by a user on the grayscale image, and is able to produce visually pleasing colorization results promptly after providing the color information.
Abstract: In this paper we present a novel colorization scheme that takes advantage of the modified morphological distance transform to propagate the color, scribbled by a user on the grayscale image First, based on the scribbled image, the topological distance values are computed for each image pixel, describing its distance to the inserted color markers These values are then complemented with the structural information and luminance changes derived from the original grayscale image The distances are then used along with gradient based features to reproduce original image structures while propagating the new colors obtained during the additive color blending process Extensive experiments performed on various kinds of natural images demonstrated the effectiveness of the proposed colorization method They also showed that the main advantage of the presented algorithm is its computational speed and ability to produce visually pleasing colorization results promptly after providing the color information