Skeleton Subsets Research Articles

Multi‐temporal scale aggregation refinement graph convolutional network for skeleton‐based action recognition

AbstractSkeleton‐based human action recognition is gaining significant attention and finding widespread application in various fields, such as virtual reality and human‐computer interaction systems. Recent studies have highlighted the effectiveness of graph convolutional network (GCN) based methods in this task, leading to a remarkable improvement in prediction accuracy. However, most GCN‐based methods overlook the varying contributions of self, centripetal and centrifugal subsets. Besides, only a single‐scale temporal feature is adopted, and the multi‐temporal scale information is ignored. To this end, firstly, in order to differentiate the importance of different skeleton subsets, we develop a refinement graph convolution, which can adaptively learn a weight for each subset feature. Secondly, a multi‐temporal scale aggregation module is proposed to extract more discriminative temporal dynamic information. Furthermore, a multi‐temporal scale aggregation refinement graph convolutional network (MTSA‐RGCN) is proposed, and four‐stream structure is also adopted in this paper, which can comprehensively model complementary features and eventually achieves a significant performance boost. In the empirical experiments, the performance of our approach has been greatly improved on both NTU‐RGB+D 60 and NTU‐RGB+D 120 datasets, compared to other state‐of‐the‐art methods.

An Iterative Morphological Decomposition Algorithm for Reduction Of Skeleton Points

Shape representation is an important aspect in image processing and computer vision. There are several skeleton transforms that lead to morphological shape representation algorithm. One of the main problems with these algorithms is in selecting the skeleton points that represent the shape component. If the numbers of skeleton subsets are reduced then the reconstruction process will be easy and time consuming. The present paper proposes a skeleton scheme that selects skeleton points based on the largest shape element. By this, overall skeleton subsets will be reduced. The present method is applied on various images and is compared with generalized skeleton transform and octagon-generating decomposition algorithm.

Skeleton Primitive Extraction Method on Textures with Different Nonlinear Wavelets

Problem Statement: A novel method for dominant skeleton extraction of textures using different nonlinear wavelet transforms is proposed in this study. In the present study 3◊3 masks are used for extraction of skeleton primitives. For a 3 ◊3 skeleton primitive there will be 2 9 skeleton primitive combinations. But the present study consi dered a skeleton primitive for the skeletonization purpose if and only if its center pixel is one and skeleton primitives are represented by the corresponding skeleton primitive weight. By this, t here will be 2 8 combinations of skeleton primitives. Approach : The skeleton primitive are used for evaluating sk eleton points. The skeleton of an object has the property that it is reduced to one point wh en the skeleton primitive used for the skeletonizat ion is exactly homothetic to the object. The dominant s keleton subset is evaluated by counting the skeleto n points. The skeleton subset that leads to the least skeleton points will be the resultant skeleton sub set. The present study classified the textures based on two methods. In the first method textures are classified based on skeleton primitive weight, whic h is nothing but based on skeleton primitive combination. In the second method classification is made based on distance function of skeleton points. Results : The proposed method is applied on 24 Brodatz text ures using the three nonlinear wavelet transformed textures. By this the dominant skeleton primitive weight is obtained for each texture. Based on the number of skeleton points dis tance measures are calculated based on which texture classification is obtained. Conclusions : The two methods are applied on Brodatz textures u sing different nonlinear wavelet transforms which classi fied the textures. The first method is appropriate if one need to classify based on skeleton subsets. The second method is appropriate if the classification is to be done based on least number of skeleton points .

Iterative generalization of morphological skeleton

This paper addresses the representation of binary images using mathematical morphology, a nonlinear theory for image processing, based on set theory. The new image representation, called “five-step” skeleton representation, is an extension of the morphological binary skeleton. It consists of calculating the morphological digital binary skeleton using squares or crosses and then reiterating the above procedure for skeleton subsets using lines (horizontal, vertical, 45° and 135°). The theoretical background of the new morphological iterative image representation is presented. Applications and results are illustrated by computer simulations.

A generalized discrete morphological skeleton transform with multiple structuring elements for the extraction of structural shape components

A common problem shared by several leading morphological shape representation algorithms is that there is much overlapping among the representative disks of the same size. A shape component represented by a group of connected disk centers sometimes uses many heavily overlapping representative disks to represent a relatively simple shape part. A shape component may also contain a large number of representative disks that form a complicated structure. We introduce a generalized discrete morphological skeleton transform that uses eight structuring elements to generate skeleton subsets so that no two skeletal points from the same skeleton subset are adjacent to each other. Each skeletal point represents a shape part that is in general an octagon with four pairs of parallel opposing sides. The number of representative points needed to represent a given shape is significantly lower than that in the standard skeleton transform. A collection of shape components needed to build a structural representation is easily derived from the generalized skeleton transform. Each shape component covers a significant area of the given shape and severe overlapping is avoided. The given shape can also be accurately approximated using a small number of shape components.

Hierarchical decomposition of multiscale skeletons

The paper presents a novel procedure to hierarchically decompose a multiscale discrete skeleton. The skeleton is a linear pattern representation that is generally recognized as a good shape descriptor. For discrete images, the discrete skeleton is often preferable. Multiresolution representations are convenient for many image analysis tasks. Our resulting skeleton decomposition shows two different types of hierarchy. The first type of hierarchy is one of different scales, as the original pattern is converted into an AND-pyramid and the skeleton is computed for each resolution level. The second type of hierarchy is established at each level of the pyramid by identifying and ranking skeleton subsets according to their permanence, where permanence is a property intrinsically related to local pattern thickness. To achieve the decomposition, both bottom-up and top-down analysis in the sense of moving from higher to lower resolution and vice versa are used. The bottom-up analysis is used to ensure that a part of the skeleton that is connected at a higher resolution level is also connected (if at all present) in the next, lower resolution level. The top-down analysis is used to build the permanence hierarchy ranking the skeleton components. Our procedure is based on the use of (3/spl times/3) local operations in digital images, so it is fast and easy to implement. This skeleton decomposition procedure is most effective on patterns having different thickness in different regions. A number of examples of decompositions of multiscale skeletons (with and without loops) are shown. The skeletons are, in most cases, nicely decomposed into meaningful parts. The procedure is general and not limited to any specific application.

HIERARCHICAL DECOMPOSITION OF DISTANCE LABELED SKELETONS

The distance labeled skeleton of a two-dimensional digital object is hierarchically decomposed into loops, branches and concatenations of branches. Decomposition is obtained by using an iterated tracing-and-deleting process alternately applied to more and more internal skeleton subsets. Every skeleton subset is associated a feature, called the relevance, which is related to the spatial extension of the object region represented by that subset. The relevance is used to decide on how to perform concatenations, as well as to rank skeleton decomposition components ascribed to the same hierarchy level. The decomposition is stable under object rotation and can be used in a recognition process accomplished via graph matching.

Parts of planar shapes

An algorithm for partitioning the shape of a planar pattern into near-convex parts is described, which utilizes ancillary processes such as dominant point detection and label propagation. The shape features taken into account are negative minima of curvature on the contour and spines, i.e. suitable skeleton subsets. Each spine has its own view of the pattern and exerts an influence over the neighboring pixels as far as no conflicts with other spines arise. The parts of the decomposition are found as the zones of influence of the spines. To favor a more compact description, the algorithm also includes a phase devoted to the extraction of loop components in correspondence with the holes of the pattern.

Morphological decomposition of 2-D binary shapes into conditionally maximal convex polygons

Mathematical morphology is an approach to image analysis based on the geometric concept of shape and size. Naturally, it provides a very effective tool for shape analysis. In this paper we present a morphological shape segmentation algorithm that decomposes a 2-D (two-dimensional) binary shape into a collection of restricted convex polygons. We use simple morphological operations to extract shape information from different skeleton subsets to form a group of conditionally maximal convex polygonal shape components. Then a set of more meaningful and natural convex polygonal components are constructed from these maximal components. The resulting algorithm is very simple and the decomposition is always unique. The components produced have well-defined mathematical characterizations and they seem to be in good agreement with the nature structures of the given shapes. The shape segments produced can be used to construct structural shape description and for other shape analysis purposes.

Boundary-constrained morphological skeleton minimization and skeleton reconstruction

A new algorithm for minimizing a morphological skeleton entitled boundary-constrained skeleton minimization (BCSM), as well as a new algorithm for reconstructing an original image from its minimized skeletal structure termed boundary-constrained skeleton reconstruction (BCSR), are proposed. The new algorithms are shown to reduce data storage requirements from (N+1) binary images represented as separate skeleton subsets with their corresponding indices, to 2 binary images composed of a binary morphological skeleton and its corresponding morphological boundary structure. In addition to a reduction in memory storage, BCSM and BCSR result in substantial savings in computational complexity. The proposed algorithms are evaluated in the context of image analysis and coding, and their performance is compared to previous algorithms proposed by Serra (1982, 1988) and by Maragos and Schafer (1986). Sample evaluations indicate a greater than 22-fold savings in computational requirements and 11-fold reduction in memory requirements. >

An efficient method for obtaining morphological skeletons

A new method using gray-scale morphological erosion is proposed to obtain the morphological skeleton of a discrete binary image. The full morphological skeleton is the union of a finite number of skeleton subsets. We also show that gray-scale erosion in this method can be replaced by a simpler operation to further improve computation time. Two efficient algorithms based on the proposed approach are presented.

Morphological skeleton representation and coding of binary images

This paper presents the results of a study on the use of morphological set operations to represent and encode a discrete binary image by parts of its skeleton, a thinned version of the image containing complete information about its shape and size. Using morphological erosions and openings, a finite image can be uniquely decomposed into a finite number of skeleton subsets and then the image can be exactly reconstructed by dilating the skeleton subsets. The morphological skeleton is shown to unify many previous approaches to skeletonization, and some of its theoretical properties are investigated. Fast algorithms that reduce the original quadratic complexity to linear are developed for skeleton decomposition and reconstruction. Partial reconstructions of the image are quantified through the omission of subsets of skeleton points. The concepts of a globally and locally minimal skeleton are introduced and fast algorithms are developed for obtaining minimal skeletons. For images containing blobs and large areas, the skeleton subsets are much thinner than the original image. Therefore, encoding of the skeleton information results in lower information rates than optimum block-Huffman or optimum runlength-Huffman coding of the original image. The highest level of image compression was obtained by using Elias coding of the skeleton.