Maximum likelihood thresholding based on population mixture models

doi:10.1016/0031-3203(92)90024-D

Pattern Recognition

Volume 25, Issue 10, October 1992, Pages 1231-1240

https://doi.org/10.1016/0031-3203(92)90024-D Get rights and content

Abstract

Maximum likelihood thresholding methods are presented on the basis of population mixture models. It turns out that the standard thresholding proposed by Otsu, which is based on a discriminant criterion and also minimizes the mean square errors between the original image and the resultant binary image, is equivalent to the maximization of the likelihood of the conditional distribution in the population mixture model under the assumption of normal distributions with a common variance. It is also shown that Kittler and Illingworth's thresholding, which minimizes a criterion related to the average classification error rate assuming normal distribution with different variances, is equivalent to the maximization of the likelihood of the joint distribution in the population mixture model. A multi-thresholding algorithm based on Dynamic Programming is also presented.

References (9)

J.S. Weszka
A survey of threshold selection techniques
Comput. Graphics Image Process.
(1978)
P.K. Sahoo et al.
A survey of thresholding techniques
Comput. Vision Graphics Image Process.
(1988)
J. Kittler et al.
Minimum error thresholding
Pattern Recognition
(1986)
N. Otsu
A threshold selection method from gray-level histograms
IEEE Trans. Syst. Man Cybern. SMC-9
(1979)

There are more references available in the full text version of this article.

Cited by (240)

Integrated digital image processing techniques and deep learning approaches for wheat stripe rust disease detection and grading
2023, Decision Analytics Journal
This study proposes integrated image processing and deep learning approaches for wheat rust disease detection and grading. Wheat stripe rust is one of the diseases affecting the crop in nearly all wheat-growing countries, where humidity and high temperatures characterize the growing season. The early and prompt detection of this disease is essential for establishing timely remedial measures to contain its spread. We present several image-processing and deep-learning approaches for detecting wheat stripe rust. The image processing techniques used in this research include single-band and dual-band processing, computing the Visual Atmospheric Resistance Index (VARI), image segmentation, and image cropping. These techniques are evaluated with different deep learning approaches like transfer learning and other Convolutional Neural Networks (CNNs) architectures. The results obtained on the dataset show that processing the images to eliminate the background as much as possible before training a suitable deep learning model with a low number of epochs optimally classifies the wheat stripe rust into the target classes (i.e., healthy, resistant, and susceptible) with an overall accuracy of 84.1% on the holdout test set. Besides the comparative analysis of various image processing approaches with different deep learning models, this study suggests multiple future research directions.
A simplified method for water depth mapping over crops during flood based on Copernicus and DTM open data
2022, Agricultural Water Management
After an extreme rainy event agricultural fields can be submerged by water. Stagnant water can be generated by river’ flooding or by soil saturation causing different damage level to crops. In this work, the flood event occurred on 3rd October 2020 in NW Italy along the Sesia river was assessed with special concern about damages affecting rice crop fields. A method was proposed aimed at detecting flooded areas and giving an estimate of water depth (WD) based on free available Copernicus data (Sentinel-1 and Sentinel-2) and digital terrain model (DTM). In particular, Sentinel-1 pre- and post-event images were compared by differencing (ΔVV). ΔVV was processed at pixel level to detect submerged areas through the thresholding Otsu’s method. A simplified morphological analysis was then performed by DTM tessellation to map WD. A further step aimed at classifying submerged areas was achieved based on DTM and a proximity analysis, making possible to separate areas where water was related to soil saturation from areas where water was coming from the river. Corine Land Cover 2018 level-3 and NDVI from a Sentinel-2 pre-event image were used to map crops that were still to be harvested at the time of flood. These were the ones that were considered while estimating the potential economic loss. A total of 255 ha of rice that still to be harvested were submerged but only 211 ha were affected by river overflow. Using local rice yield and price the resulting economic loss was about 2,200,000 €.
Experimental and theoretical investigation on the effect of inclined surface on pool fire behavior
2022, Process Safety and Environmental Protection
Potential pool fire hazards due to the leakage of combustible liquid chemical materials during storage and transportation pose a threat to process safety in the process industry. A pool fire may occur on the inclined surface of broken pipelines or facilities. The influence of an inclined surface on pool fire behavior was experimentally and theoretically studied. The results showed that the flame tilt angle and flame length increased with inclination angle, while the flame height exhibited an opposite trend. These phenomena were mainly caused by the restriction of air entrainment due to the inclined surface. The flame pulsation frequency generally decreased with the increase of inclination angle and was weakly dependent on the heat release rate. Dimensionless models were developed to predict the flame tilt angle, flame length, flame height, and flame pulsation frequency. The predictive values were in good agreement with experimental data.
Multithresholding techniques in SAR image classification
2021, Remote Sensing Applications: Society and Environment
Beyond bilevel segmentation, multithresholding (MT) is a powerful technique that is seldom considered in image analysis. In active imaging techniques, however, MT appears as a significant alternative due to its obliviousness to the inherent nonlinear noise present in these kind of images. This is especially the case in satellite synthetic aperture radar (SAR) images, which are becoming an increasingly popular information source in remote sensing, given that the acquisition is independent of weather and daylight conditions. In SAR images, data-dependent multiplicative noise (speckle) hampers most of the image filtering techniques available from linear theory, and thus nonlinear techniques like MT appear to be promising. There are, however, a large amount of proposals in this direction, each with different theoretical or empirical justifications, and thus a careful analysis of their advantages in results quality and computational cost have to be assessed. In this paper we survey the most representative MT techniques and methods applied to SAR imagery (both synthetic and actual satellite images), and evaluate them in terms of region segmentation quality and computational cost. Results show that the maximum likelihood method provides the best quality segmentation results at the expense of higher computation times, while a state transition based method provides the fastest results with acceptable quality, and that all methods can be assessed with a simple tradeoff representation.
Detail reconstruction in binary ghost imaging by using point-by-point method
2021, Optics Communications
We propose a new local-binary ghost imaging by using point-by-point method. This method can compensate the degradation of imaging quality due to the loss of information during binarization process. The numerical and experimental results show that the target details can be reconstructed well by this method when compared with traditional ghost imaging. By comparing the differences of the speckle patterns from different binarization methods, we also give the corresponding explanation. Our results may have the potential applications in areas with high requirements for imaging details, such as target recognition.
Automated detection of damaged buildings in post-disaster scenarios: a case study of Kahramanmaraş (Türkiye) earthquakes on February 6, 2023
2023, Natural Hazards

View all citing articles on Scopus

View full text

Maximum likelihood thresholding based on population mixture models

Abstract

Comput. Graphics Image Process.

Comput. Vision Graphics Image Process.

Pattern Recognition

A threshold selection method from gray-level histograms

IEEE Trans. Syst. Man Cybern. SMC-9