Feature Guided Biased Gaussian Mixture Model for image matching

Sun, Kun; Li, Peiran; Tao, Wenbing; Tang, Yuanyan

doi:10.1016/j.ins.2014.10.029

Cited by 20 publications

(5 citation statements)

References 55 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…The straightforward consensus algorithm, MV, has been thoroughly studied [15] [32]. Besides MV, many consensus algorithms utilize the inference of a useful probabilistic model for estimating latent variables [1] [39]. The classic approach, DS [5] proposed by Dawid and Skene, models a confusion matrix of each annotator and the class prior and uses an EM procedure to infer the estimated labels of examples.…”

Section: Related Workmentioning

confidence: 99%

Consensus algorithms for biased labeling in crowdsourcing

Zhang

Sheng

et al. 2017

Information Sciences

View full text Add to dashboard Cite

Although it has become an accepted lay view that when labeling objects through crowdsourcing systems, non-expert annotators often exhibit biases, this argument lacks sufficient evidential observation and systematic empirical study. This paper initially analyzes eight real-world datasets from different domains whose class labels were collected from crowdsourcing systems. Our analyses show that biased labeling is a systematic tendency for binary categorization; in other words, for a large number of annotators, their labeling qualities on the negative class (supposed to be the majority) are significantly greater than are those on the positive class (minority). Therefore, the paper empirically studies the performance of four existing EM-based consensus algorithms, DS, GLAD, RY, and ZenCrowd, on these datasets. Our investigation shows that all of these state-of-the-art algorithms ignore the potential bias characteristics of datasets and perform badly although they model the complexity of the systems. To address the issue of handling biased labeling, the paper further proposes a novel consensus algorithm, namely adaptive weighted majority voting (AWMV), based on the statistical difference between the labeling qualities of the two classes. AWMV utilizes the frequency of positive labels in the multiple noisy label set of each example to obtain a bias rate and then assigns weights derived from the bias rate to negative and positive labels. Comparison results among the five consensus algorithms (AWMV and the four existing) show that the proposed AWMV algorithm has the best overall performance. Finally, this paper notes some potential related topics for future study.

show abstract

Section: Related Workmentioning

confidence: 99%

Consensus algorithms for biased labeling in crowdsourcing

Zhang

Sheng

et al. 2017

Information Sciences

View full text Add to dashboard Cite

show abstract

“…We can use any existing maximum clique-searching algorithm to solve Equation (20). As far as we know, a recent MC algorithm proposed in [38] can efficiently solve the maximum clique problem, which combines tree searching with efficient bounding and pruning based on graph coloring.…”

Section: Graph Construction and Maximum Clique Algorithmmentioning

confidence: 99%

“…In general, feature-based methods mainly contain two steps: feature extraction and feature matching. In recent years, some feature-based methods [13][14][15][16][17][18][19][20][21][22] have been proposed. These algorithms have some common steps including keypoint detection, keypoint description, and keypoint matching.…”

Section: Introductionmentioning

confidence: 99%

EMCM: A Novel Binary Edge-Feature-Based Maximum Clique Framework for Multispectral Image Matching

Fang

et al. 2019

Remote Sensing

View full text Add to dashboard Cite

Seeking reliable correspondence between multispectral images is a fundamental and important task in computer vision. To overcome the nonlinearity problem occurring in multispectral image matching, a novel, edge-feature-based maximum clique-matching frame (EMCM) is proposed, which contains three main parts: (1) a novel strong edge binary feature descriptor, (2) a new correspondence-ranking algorithm based on keypoint distinctiveness analysis algorithms in the feature space of the graph, and (3) a false match removal algorithm based on maximum clique searching in the correspondence space of the graph considering both position and angle consistency. Extensive experiments are conducted on two standard multispectral image datasets with respect to the three parts. The feature-matching experiments suggest that the proposed feature descriptor is of high descriptiveness, robustness, and efficiency. The correspondence-ranking experiments validate the superiority of our correspondences-ranking algorithm over the nearest neighbor algorithm, and the coarse registration experiments show the robustness of EMCM with varied interferences.

show abstract

“…In the shallow layers of the network, it is hard to learn appropriate weight because the features in these layers are less recognizable. In fact, Lowe Ratio (Lowe 2004), i.e., the side information generated during feature matching, is proved to be powerful prior information to determine the confidence of each point being inlier (Goshen and Shimshoni 2008;Brahmachari and Sarkar 2009;Sun et al 2015;Tao and Sun 2014). Based on this observation, we propose a Bayesian attentive context normalization (BACN) to mine prior information for better reducing the noise of outliers to global context.…”

Section: Introductionmentioning

confidence: 99%

Cascade Network with Guided Loss and Hybrid Attention for Finding Good Correspondences

Chen

Yang

Tao

2021

AAAI

Self Cite

View full text Add to dashboard Cite

Finding good correspondences is a critical prerequisite in many feature based tasks. Given a putative correspondence set of an image pair, we propose a neural network which finds correct correspondences by a binary-class classifier and estimates relative pose through classified correspondences. First, we analyze that due to the imbalance in the number of correct and wrong correspondences, the loss function has a great impact on the classification results. Thus, we propose a new Guided Loss that can directly use evaluation criterion (Fn-measure) as guidance to dynamically adjust the objective function during training. We theoretically prove that the perfect negative correlation between the Guided Loss and Fn-measure, so that the network is always trained towards the direction of increasing Fn-measure to maximize it. We then propose a hybrid attention block to extract feature, which integrates the Bayesian attentive context normalization (BACN) and channel-wise attention (CA). BACN can mine the prior information to better exploit global context and CA can capture complex channel context to enhance the channel awareness of the network. Finally, based on our Guided Loss and hybrid attention block, a cascade network is designed to gradually optimize the result for more superior performance. Experiments have shown that our network achieves the state-of-the-art performance on benchmark datasets. Our code will be available in https://github.com/wenbingtao/GLHA.

show abstract

Feature Guided Biased Gaussian Mixture Model for image matching

Cited by 20 publications

References 55 publications

Consensus algorithms for biased labeling in crowdsourcing

Consensus algorithms for biased labeling in crowdsourcing

EMCM: A Novel Binary Edge-Feature-Based Maximum Clique Framework for Multispectral Image Matching

Cascade Network with Guided Loss and Hybrid Attention for Finding Good Correspondences

Contact Info

Product

Resources

About