Shape Similarity Analysis by Self-Tuning Locally Constrained Mixed-Diffusion

Abstract

Authors and/or their employers shall have the right to post the accepted version of IEEE-copyrighted articles on their own personal servers or the servers of their institutions or employers without permission from IEEE, provided that the posted version includes a prominently displayed IEEE copyright notice (as shown in 8. Abstract-Similarity analysis is a powerful tool for shape matching/retrieval and other computer vision tasks. In the literature, various shape (dis)similarity measures have been introduced. Different measures specialize on different aspects of the data. In this paper, we consider the problem of improving retrieval accuracy by systematically fusing several different measures. To this end, we propose the locally constrained mixeddiffusion method, which partly fuses the given measures into one and propagates on the resulted locally dense data space. Furthermore, we advocate the use of self-adaptive neighborhoods to automatically determine the appropriate size of the neighborhoods in the diffusion process, with which the retrieval performance is comparable to the best manually tuned kNNs. The superiority of our approach is empirically demonstrated on both shape and image datasets. Our approach achieves a score of 100% in the bull's eye test on the MPEG-7 shape dataset, which is the best reported result to date.Index Terms-Shape similarity analysis, shape/image retrieval, locally constrained mixed-diffusion.

A unified model for accelerating unsupervised iterative re‐ranking algorithms

“…However, these approaches fail to return satisfactory results in many situations, mainly due to the well-known semantic gap problem [21][22][23]. This has motivated research attempts to improve distance metrics in CBIR systems in the past few years [2,[7][8][9][10][11][12][13]24], leading to promising results considering several approaches and post-processing techniques [25][26][27][28].…”

“…Since labeling is often a laborious and timeconsuming task, whereas it is far easier to obtain unlabeled data, these techniques often represent a very attractive solution. Additionally, we adopted an iterative strategy to process contextual information with the goal of re-ranking the images returned at top positions of ranked lists [3,[10][11][12]30].…”

“…As the effectiveness of these systems is very dependent on the distance metrics adopted, substantial work [2,[7][8][9][10][11][12][13] has also been conducted with the aim of improving the computation of these metrics.…”

Dynamic Re-ranking with Deep Features Fusion for Person Re-identification

“…Compared to the traditional pairwise affinity which measures the similarity between two data points, the diffusion process explores the affinity information among all data points in a global view. In addition, the manifold ranking can be ranged into two principal types, diffusion on a single graph [8,36,37] and on multiple graphs [15,30,35,42]. The latter is designed for features fusion and metrics fusion, named as fusion by diffusion.…”

Maximum Variance Hashing via Column Generation

“…Nearest neighbor search is a fundamental problem in many applications concerned with information retrieval, including content-based multimedia retrieval [1][2][3], object and scene recognition [4], and image matching [5]. Due to the exciting advancement of data acquisition techniques, more and more data have been produced in recent years, leading these applications to suffer from the expensive time and storage demand.…”