B. Uma Shankar

Segmentation of multispectral remote sensing images using active support vector machines

The problem of scarcity of labeled pixels, required for segmentation of remotely sensed satellite images in supervised pixel classification framework, is addressed in this article. A support vector machine (SVM) is considered for classifying the pixels into different landcover types. It is initially designed using a small set of labeled points, and subsequently refined by actively querying for the labels of pixels from a pool of unlabeled data. The label of the most interesting/ ambiguous unlabeled point is queried at each step. Here, active learning is exploited to minimize the number of labeled data used by the SVM classifier by several orders. These features are demonstrated on an IRS-1A four band multi-spectral image. Comparison with related methods is made in terms of number of data points used, computational time and a cluster quality measure.

Segmentation of remotely sensed images with fuzzy thresholding, and quantitative evaluation

Effectiveness of various fuzzy thresholding techniques (based on entropy of fuzzy sets, fuzzy geometrical properties, and fuzzy correlation) is demonstrated on remotely sensed (IRS and SPOT) images. A new quantitative index for image segmentation using the concept of homogeneity within regions is defined. Results are compared with those of probabilistic thresholding, and fuzzy c-means and hard c-means clustering algorithms, both in terms of index value (quantitatively) and structural details (qualitatively). Fuzzy set theoretic algorithms are seen to be superior to their respective non-fuzzy counterparts. Among all the techniques, fuzzy correlation, followed by fuzzy entropy, performed better for extracting the structures. Fuzzy geometry based thresholding algorithms produced a single stable threshold for a wide range of membership variation.

Robust Radiomics Feature Quantification Using Semiautomatic Volumetric Segmentation

Due to advances in the acquisition and analysis of medical imaging, it is currently possible to quantify the tumor phenotype. The emerging field of Radiomics addresses this issue by converting medical images into minable data by extracting a large number of quantitative imaging features. One of the main challenges of Radiomics is tumor segmentation. Where manual delineation is time consuming and prone to inter-observer variability, it has been shown that semi-automated approaches are fast and reduce inter-observer variability. In this study, a semiautomatic region growing volumetric segmentation algorithm, implemented in the free and publicly available 3D-Slicer platform, was investigated in terms of its robustness for quantitative imaging feature extraction. Fifty-six 3D-radiomic features, quantifying phenotypic differences based on tumor intensity, shape and texture, were extracted from the computed tomography images of twenty lung cancer patients. These radiomic features were derived from the 3D-tumor volumes defined by three independent observers twice using 3D-Slicer, and compared to manual slice-by-slice delineations of five independent physicians in terms of intra-class correlation coefficient (ICC) and feature range. Radiomic features extracted from 3D-Slicer segmentations had significantly higher reproducibility (ICC = 0.85±0.15, p = 0.0009) compared to the features extracted from the manual segmentations (ICC = 0.77±0.17). Furthermore, we found that features extracted from 3D-Slicer segmentations were more robust, as the range was significantly smaller across observers (p = 3.819e-07), and overlapping with the feature ranges extracted from manual contouring (boundary lower: p = 0.007, higher: p = 5.863e-06). Our results show that 3D-Slicer segmented tumor volumes provide a better alternative to the manual delineation for feature quantification, as they yield more reproducible imaging descriptors. Therefore, 3D-Slicer can be employed for quantitative image feature extraction and image data mining research in large patient cohorts.

A novel approach to neuro-fuzzy classification

A new model for neuro-fuzzy (NF) classification systems is proposed. The motivation is to utilize the feature-wise degree of belonging of patterns to all classes that are obtained through a fuzzification process. A fuzzification process generates a membership matrix having total number of elements equal to the product of the number of features and classes present in the data set. These matrix elements are the input to neural networks. The effectiveness of the proposed model is established with four benchmark data sets (completely labeled) and two remote sensing images (partially labeled). Different performance measures such as misclassification, classification accuracy and kappa index of agreement for completely labeled data sets, and beta index of homogeneity and Davies-Bouldin (DB) index of compactness for remotely sensed images are used for quantitative analysis of results. All these measures supported the superiority of the proposed NF classification model. The proposed model learns well even with a lower percentage of training data that makes the system fast.

A novel fuzzy classifier based on product aggregation operator

The present article proposes a fuzzy set-based classifier with a better learning and generalization capability. The proposed classifier exploits the feature-wise degree of belonging of a pattern to all classes, generalization in the fuzzification process and the combined class-wise contribution of features effectively. The classifier uses a @p-type membership function and product aggregation reasoning rule (operator). Its effectiveness is verified with two conventional (completely labeled) data sets and two remote sensing images (partially labeled data sets). The proposed classifier is compared with similar fuzzy methods. Different performance measures are used for quantitative evaluation of the proposed classifier.

Wavelet-fuzzy hybridization: Feature-extraction and land-cover classification of remote sensing images

Abstract: A wavelet feature based supervised scheme for fuzzy classification of land covers in multispectral remote sensing images is proposed. The proposed scheme is developed in the framework of wavelet-fuzzy hybridization, a soft computing approach. The wavelet features obtained from wavelet transform on an image provides spatial and spectral characteristics (i.e., texture information) of pixels and hence can be utilized effectively for improving accuracy in classification, instead of using original spectral features. Four different fuzzy classifiers are considered for this purpose and evaluated using different wavelet features. Wavelet feature based fuzzy classifiers produced consistently better results compared to original spectral feature based methods on various images used in the present investigation. Further, the performance of the Biorthogonal3.3 (Bior3.3) wavelet is observed to be superior to other wavelets. This wavelet in combination with fuzzy product aggregation reasoning-rule outperformed all other methods. Potentiality of the proposed soft computing approach in isolating various land covers are evaluated both visually and quantitatively using indexes like @b measure of homogeneity and Xie-Beni measure of compactness and separability.

Neuro-fuzzy-combiner: an effective multiple classifier system

A neuro-fuzzy-combiner (NFC) is proposed to design an efficient multiple classifier system (MCS) with an aim to have an effective solution scheme for difficult classification problems. Although, a number of combiners exist in the literature, they do not provide consistently good performance on different datasets. In this scenario: 1) we propose an effective multiple classifier system (MCS) based on NFC that fuses the output of a set of fuzzy classifiers; 2) conduct an extensive experimental analysis to justify the effectiveness of the proposed NFC. In the proposed technique, we used a neural network to combine the output of a set of fuzzy classifiers using the principles of neuro-fuzzy hybridisation. The neural combiner adaptively learns its parameters depending on the input data, and thus the output is robust. Superiority of the proposed combiner has been demonstrated experimentally on five standard datasets and two remote sensing images. It performed consistently better than the existing combiners over all the considered datasets.

Fast mean filtering technique (FMFT)

This article presents a novel method for mean filtering that reduces the required number of additions and eliminates the need for division altogether. The time reduction is achieved using basic store-and-fetch operations and is irrespective of the image or neighbourhood size. This method has been tested on a variety of greyscale images and neighbourhood sizes with promising results. These results indicate that the relative time requirement reduces with increase in image size. The methods efficiency also improves significantly with increase in neighbourhood size thereby making it increasingly useful when dealing with large images.

Medical image analysis for cancer management in natural computing framework

Natural computing, through its repertoire of nature-inspired strategies, is playing a major role in the development of intelligent decision-making systems. The objective is to provide flexible, application-oriented solutions to current medical image analysis problems. It encompasses fuzzy sets, neural networks, genetic algorithms, rough sets, swarm intelligence, and a host of other paradigms, mimicking biological and physical processes from nature.Radiographic imaging modalities, like computed tomography (CT), positron emission tomography (PET), and magnetic resonance imaging (MRI), help in providing improved diagnosis, prognosis and treatment planning for cancer. This survey highlights the role of natural computing, in efficiently analyzing radiographic medical images, for improved tumor management. We also provide a categorization of the segmentation, feature extraction and selection methods, based on different natural computing technologies, with reference to the application - involving malignancy of the brain, breast, prostate, skin, lung, and liver.

Novel classification and segmentation techniques with application to remotely sensed images

The article deals with some new results of investigation, both theoretical and experimental, in the area of image classification and segmentation of remotely sensed images. The article has mainly four parts. Supervised classification is considered in the first part. The remaining three parts address the problem of unsupervised classification (segmentation). The effectiveness of an active support vector classifier that requires reduced number of additional labeled data for improved learning is demonstrated in the first part. Usefulness of various fuzzy thresholding techniques for segmentation of remote sensing images is demonstrated in the second part. A quantitative index of measuring the quality of classification/segmentation in terms of homogeneity of regions is introduced in this regard. Rough entropy (in granular computing framework) of images is defined and used for segmentation in the third part. In the fourth part a homogeneous region in an image is defined as a union of homogeneous line segments for image segmentation. Here Hough transform is used to generate these line segments. Comparative study is also made with related techniques.