China
Africa
Explore chapters and articles related to this topic
Iris Segmentation in the Wild Using Encoder-Decoder-Based Deep Learning Techniques
Published in Gaurav Jaswal, Vivek Kanhangad, Raghavendra Ramachandra, AI and Deep Learning in Biometric Security, 2021
Shreshth Saini, Divij Gupta, Ranjeet Ranjan Jha, Gaurav Jaswal, Aditya Nigam
The authors of Ref. [48] proposed a multi-stage technique. First, a moving window of circular shape was used for the pupil estimation, following which the estimation of the pupil was done through the standard-deviation peaks in both x as well as y directions, and after that, a median-filter reduced the eyelash effects. In Ref. [3], the authors proposed AdaBoost for eye detection for further segmentation. Reference [80] presents an unsupervised approach where images were modelled as Markov random field. Graph-cut method extracted the texture region, and for the iris segmentation image, intensities were exploited. Roy et al. [88] proposed a non-ideal iris recognition method, in which they used a Mumford-Shah segmentation method. All these classical approaches claim to handle various noises, distortion, and non-ideal iris images, but all being rule-based feature-driven approaches are limited in handling the variation of a non-ideal iris image. In Table 12.2, some classical approaches are compared based on their novelty and performance.
Hybrid Fuzzy Classifiers
Published in Anil Kumar, Priyadarshi Upadhyay, A. Senthil Kumar, Fuzzy Machine Learning Algorithms for Remote Sensing Image Classification, 2020
Anil Kumar, Priyadarshi Upadhyay, A. Senthil Kumar
In this chapter, we describe fuzzy hybridization techniques wherein we use different properties of local features present in the image for improved classification applications. The properties of local features can be measured with a variety of functions. Some of the most common features are described as below: Markov random field (MRF) is an approach widely used for characterizing contextual information. Contextual can be understood as correlation, and in this chapter spatial correction has been considered.Entropy in information technology defines uncertainty, and this uncertainty concept has been applied to study the effect of entropy during classification.Similarity/dissimilarity measures with many types of distance norms and can be applied in classifiers using distance function.Spectral information divergence (SID) is a random probability distribution measure of classification that matches pixels in the image to the reference spectra, and for doing so, it utilizes the divergence function.Spectral angle mapper (SAM) based spectral classifier uses angles to match unknown pixels to reference spectra.
Probability Models of Data Generation
Published in Richard M. Golden, Statistical Machine Learning, 2020
Markov random fields are widely used in image processing probabilistic models. Consider an image processing problem where each pixel in an image is represented as a random variable whose value is an intensity measurement. Assume the local conditional probability density of the ith pixel random variable is functionally dependent upon the values of the ith pixel random variable values surrounding the ith pixel random variable. For example, a typical local constraint might be that the probability the ith pixel random variable value takes on the intensity level of black will be larger when the values of the pixel random variables surrounding the ith pixel also have values corresponding to darker intensity values. Notice this is an example of a homogeneous Markov random field since the parametric form of the local conditional density for the ith pixel random variable in the field is exactly the same for all other pixel random variables in the image.
IM-GSO: A Community Directed Group Search Optimization Approach for Influence Maximization
Published in Cybernetics and Systems, 2018
Influence maximization is a fundamental problem in complex networks analytics. This problem was first analyzed to be an algorithmic problem by Domingos and Richardson (2001) in the domain of viral marketing by modeling market place as a network of customers. They used market value of customers to determine most trusted target customers who can spread the companies influence the most, in order to maximize the profit by minimizing the budget. The network was modeled as Markov random field by using probabilistic modeling. This problem was later analyzed to be NP hard by Kempe et al. (2003) who then defined the influence maximization problem as a discrete optimization problem. Kempe et al. (2003) proposed a greedy approximation algorithm to obtain a solution within 63% of optimal for varied information diffusion models.
Jump regression, image processing, and quality control
Published in Quality Engineering, 2018
In the image processing literature, people traditionally describe an image as a Markov random field (MRF). In that framework, observed image intensities of an image are assumed to have the Markov property that the observed intensity at a given pixel depends only on the observed intensities in a neighborhood of the given pixel. It has been proved that the random field has a Gibbs distribution in that case (e.g., Besag 1974). If the true image intensities are assumed to have a prior distribution that is also a Gibbs distribution, then the posterior distribution of the true image would be a Gibbs distribution too and the true image can be estimated by the maximum a posteriori (MAP) estimator (cf., Geman and Geman 1984). Many image processing methods were developed in that framework.
An underwater lighting and turbidity image repository for analysing the performance of image-based non-destructive techniques
Published in Structure and Infrastructure Engineering, 2018
Michael O’Byrne, Franck Schoefs, Vikram Pakrashi, Bidisha Ghosh
This section compares the performance of three distinct types of stereo correspondence algorithms. The first is a Belief Propagation (BP) Markov Random Field (MRF) method (Sun, Shum, & Zheng, 2002), which is a hierarchical (coarse-to-fine) algorithm that operates on an image pyramid, where results from coarser levels are used to constrain a more local search at finer levels. The MRF model takes into account the differences between pixel intensity values between corresponding points and the spatial relationship between the horizontal disparities. The disparity relates how far each point is from the camera, i.e. its depth in the scene. The goal is to find a piecewise smooth horizontal disparity map consistent with the observed data which minimises total energy. This method promotes a smooth disparity map as it penalises cases where neighbouring pixels have different disparity values.