A Bayesian Approach to Image Recognition Based on Separable Lattice Hidden Markov Models

Sawada, Kei; Tamamori, Akira; Hashimoto, Kei; Nankaku, Yoshihiko; Tokuda, Keiichi

doi:10.1587/transinf.2016edp7112

Cited by 6 publications

(5 citation statements)

References 33 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…9 in Dataset2 was modeled by using three factors and the tuning parameter τ = 1/2000 in order to visualize model parameters. Figure 10 shows the values of the mean vector μ k in (36), and the eigen images W k in (34) were represented in grayscale. From Fig.…”

Section: Resultsmentioning

confidence: 99%

“…Figure shows some examples of images for the experiments. Also, 40 × 40 HMM states were used for the experiment, because our prior work and preliminary experiments showed the highest recognition performance in 40 × 40 HMM states.…”

Section: Methodsmentioning

confidence: 99%

“…The details of VB E‐ and M‐step are given in Appendix A. In the proposed method, HMEMs that have the same number of states as the number of pixels of the input images are equivalent to the conventional FA using the VB method , and HMEMs with a zero factor become the SL‐HMMs using the VB method .…”

Section: Bayesian Framework For Training Of Hmemsmentioning

confidence: 99%

See 2 more Smart Citations

A bayesian framework for image recognition based on hidden markov eigen‐image models

Sawada

Hashimoto

Nankaku

et al. 2018

IEEJ Transactions Elec Engng

Self Cite

View full text Add to dashboard Cite

An image recognition method based on hidden Markov eigen‐image models (HMEMs) using a Bayesian framework is proposed and experimentally evaluated. HMEMs have been proposed as a model with two advantageous properties: invariances to the size and location of the object to be recognized, and a linear feature extraction based on statistical analysis. However, the training of HMEMs easily falls into the overfitting problem because HMEMs have a complex model structure, especially when there is insufficient training data. This study aims to accurately estimate HMEMs using the variational Bayesian (VB) method. The VB method can utilize prior distributions representing useful prior information and is expected to have a high generalization ability due to the marginalization of model parameters. Furthermore, to relax the local maximum problem in the VB method, the deterministic annealing expectation maximization algorithm is applied to train HMEMs. Experiments on face recognition indicate that the proposed method offers a significantly improved image recognition performance. Additionally, comparative experimental results show that the proposed method is more robust to geometric variations than convolutional neural networks when the amount of training data is insufficient. © 2018 Institute of Electrical Engineers of Japan. Published by John Wiley & Sons, Inc.

show abstract

Section: Resultsmentioning

confidence: 99%

Section: Methodsmentioning

confidence: 99%

See 1 more Smart Citation

A bayesian framework for image recognition based on hidden markov eigen‐image models

Sawada

Hashimoto

Nankaku

et al. 2018

IEEJ Transactions Elec Engng

Self Cite

View full text Add to dashboard Cite

show abstract

“…The EM algorithm is simple and numerically stable, but when the initial parameters deviate from the actual parameters and the algorithm converges slowly, the EM algorithm needs more iterations. Altmanetal also pointed out that when estimating the maximum probability of the estimated parameter, the maximum numerical method was faster than the EM algorithm [6][7][8]. The integration of the hidden model and the prior algorithm reduces the computation of EM algorithm and improves the efficiency.…”

Section: Improving the Introduction Of Hidden Markovmentioning

confidence: 99%

Improved Hidden Markov Algorithm Based on Bayes in Low Dose CT Images

2023

IJMC

View full text Add to dashboard Cite

In order to reduce the radiation exposure of patients during CT scan, low-dose CT images were produced, but the disadvantage was that the image quality was reduced. The Bayesian maximum posterior probability estimation (Bayesian MAP) method is an applied statistical method that can estimate the original noise-independent coefficients from the noise-contaminated image detail coefficients. This paper aims to study the application of bayesian based improved hidden markov algorithm in low-dose CT images, which has become the focus of CT research in recent years. In this experiment, hmrf-em, eHMRF algorithm and hmrf-msa-em algorithm were firstly analyzed by mathematical statistics within the experimental scope, and the superiority of this algorithm was compared by looking at different coefficients. The classification and statistical analysis of the re-data statistical method were carried out by using the naive bayesian algorithm and the improved hidden markov algorithm based on bayes. And the use of a single variable method to compare the use of bayesian based improved hidden markov algorithm in the low-dose CT image imaging whether there are different changes, and the degree of change. Experimental data show that the improved hidden markov algorithm based on bayes achieves higher values of Jaccard, Dice and CCR at different noise levels. The improved hidden markov algorithm based on bayes is clearer than the low-dose CT images obtained by the naive bayes algorithm. In various medical images, the improved hidden markov algorithm based on bayes plays an obvious role in changing the resolution of low-dose CT images. Experimental data show that compared with other denoising methods, the peak signal-to-noise ratio of the de-noised image can be improved, the detailed features of the image can be retained better, the visual effect can be improved by 46.78%, and the correct segmentation rate can reach 95.75%.

show abstract

“…On this basis, we combine the proposed sub-algorithm with other commonly used RW algorithms. 16 This unified view will make it possible to transfer internal discovery between different RW algorithms and provide new ideas for designing new RW algorithms by adding or changing secondary nodes. To verify the second advantage, we designed a new subalgorithm with tags to solve the segmentation problem of slender objects.…”

Section: Introductionmentioning

confidence: 99%

Fault diagnosis method of mechanical power system based on image processing technology

Qiu

Zhou³

2020

International Journal of Advanced Robotic Systems

View full text Add to dashboard Cite

Fault detection and diagnosis become one of today’s hot spots, which describes that image information is an important form of fault information, it can quickly through the image processing technique, and can accurately extract the characteristic signal. This article selects the color of the particle image, the integrated use of digital image processing, pattern recognition theory, the characteristic parameters of tribology knowledge, as well as the extraction, optimization, and digital; verifies the feasibility of iron spectrum of abrasive fault recognition, and provides a new efficient ferrographic wear particle image recognition method. Firstly, the grindstone image of the original color diesel engine was preprocessed, and the grindstone image of the ferrograph was identified by directly selecting grindstone from the preprocessed ferrograph image and selecting the target grindstone. According to the two types of abrasive particles, the characteristic parameters were first classified, and then the values of the characteristic parameters were obtained through the training and learning of the sample abrasive particles. In view of the large number of characteristic parameters of ferro-spectrum abrasive particles, this article determined the characteristic parameters suitable for the identification of abrasive particles in this article through the feature optimization and proved the correctness of the identification of characteristic parameters of abrasive particles through the test.

show abstract

A Bayesian Approach to Image Recognition Based on Separable Lattice Hidden Markov Models

Cited by 6 publications

References 33 publications

A bayesian framework for image recognition based on hidden markov eigen‐image models

A bayesian framework for image recognition based on hidden markov eigen‐image models

Improved Hidden Markov Algorithm Based on Bayes in Low Dose CT Images

Fault diagnosis method of mechanical power system based on image processing technology

Contact Info

Product

Resources

About