The superpixel segmentation algorithm, as a preprocessing technique, should show good performance in fast segmentation speed, accurate boundary adherence and homogeneous regularity. A fast superpixel segmentation algorithm by iterative edge refinement (IER) works well on optical images. However, it may generate poor superpixels for Polarimetric synthetic aperture radar (PolSAR) images due to the influence of strong speckle noise and many small-sized or slim regions. To solve these problems, we utilized a fast revised Wishart distance instead of Euclidean distance in the local relabeling of unstable pixels, and initialized unstable pixels as all the pixels substituted for the initial grid edge pixels in the initialization step. Then, postprocessing with the dissimilarity measure is employed to remove the generated small isolated regions as well as to preserve strong point targets. Finally, the superiority of the proposed algorithm is validated with extensive experiments on four simulated and two real-world PolSAR images from Experimental Synthetic Aperture Radar (ESAR) and Airborne Synthetic Aperture Radar (AirSAR) data sets, which demonstrate that the proposed method shows better performance with respect to several commonly used evaluation measures, even with about nine times higher computational efficiency, as well as fine boundary adherence and strong point targets preservation, compared with three state-of-the-art methods.
Clustering is a long-standing important research problem. However, it remains challenging when handling large-scale web data from different types of information resources such as user profile, comments, user preferences and so on. All these aspects can be seen as different views and often admit the same underlying clustering of the data. In this paper, we present a novel Semantic Weighted Non-negative Matrix Factorization (SWNMF) multi-view clustering framework, which can provide an efficient weighted matrix factorization framework, dexterously manipulate multi-view web content, and easily explore the sparseness problem in semantic space of data. Specifically, each view of dataset forming a huge sparse matrix, which results in the non-robust characteristic during the matrix decomposition process, and further influences the accuracy of clustering results. To address above problem, we attempt to use some preference information (e.g. rating values) given by the users as latent semantic information to handle those features that are unobserved in each data point so as to resolve the sparseness problem in all views matrices. To combine multiple views in our large corpus, the overall objective of our proposed SWNMF is to minimize the loss function of weighted non-negative matrix factorization (NMF) under the l 2,1-norm and the co-regularized constraint under the F-norm. Extensive experiments on our large-scale multi-view web datasets demonstrate the competitive performance of our solution. INDEX TERMS Multi-view clustering, semantic information, non-negative matrix factorization.
Multi-view image/video stitching algorithm is an extensive research area in computer vision and image based rendering. Most researches focus on stitching the images from different views with assumption that those images have been already aligned in temporal domain. However it is not the case in real application. If the images from different views are not aligned in temporal domain, or in another words, not time synchronized, the corresponding feature points or regions will not be located correctly among different views, which will result in ghost objects appearing in the final stitching/rendering result. In this paper, we present an epipolar geometry consistency scoring scheme to guide temporal aligned video frame pair selection for multiview video stitching application. Essentially, the proposed scheme allows us to determine whether a given pair of video frames is temporally aligned well for video stitching. Experimental results confirm that better video stitching results can be obtained with the proposed scheme in place.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.
hi@scite.ai
10624 S. Eastern Ave., Ste. A-614
Henderson, NV 89052, USA
Copyright © 2024 scite LLC. All rights reserved.
Made with 💙 for researchers
Part of the Research Solutions Family.