N. Sri Madhava Raja

Otsu based optimal multilevel image thresholding using firefly algorithm

Histogram based multilevel thresholding approach is proposed using Brownian distribution (BD) guided firefly algorithm (FA). A bounded search technique is also presented to improve the optimization accuracy with lesser search iterations. Otsus between-class variance function is maximized to obtain optimal threshold level for gray scale images. The performances of the proposed algorithm are demonstrated by considering twelve benchmark images and are compared with the existing FA algorithms such as Levy flight (LF) guided FA and random operator guided FA. The performance assessment comparison between the proposed and existing firefly algorithms is carried using prevailing parameters such as objective function, standard deviation, peak-to-signal ratio (PSNR), structural similarity (SSIM) index, and search time of CPU. The results show that BD guided FA provides better objective function, PSNR, and SSIM, whereas LF based FA provides faster convergence with relatively lower CPU time.

Robust Color Image Multi-thresholding Using Between-Class Variance and Cuckoo Search Algorithm

Multi-level image thresholding is a well known pre-processing procedure, commonly used in variety of image related domains. Segmentation process classifies the pixels of the image into various group based on the threshold level and intensity value. In this paper, colour image segmentation is proposed using Cuckoo Search (CS) algorithm. The performance of the proposed technique is validated with the Bacterial Forage Optimization (BFO) and Particle Swarm Optimization (PSO). The qualitative and quantitative investigation is carried out using the parameters, such as CPU time, between-class variance value and image quality measures, such as Mean Structural Similarity Index Matrix (MSSIM), Normalized Absolute Error (NAE), Structural Content (SC) and PSNR. The robustness of the implemented segmentation procedure is also verified using the image dataset smeared with the Gaussian Noise (GN) and Speckle Noise (SN). The study shows that, CS algorithm based multi-level segmentation offers better result compared with BFO and PSO.

Optimal Multilevel Image Thresholding to Improve the Visibility of Plasmodium sp. in Blood Smear Images

Malaria is one of the mosquito-borne communicable diseases for humans caused due to Plasmodium sp. During the treatment process, it is necessary to identify the exact Plasmodium sp. in order to give the specific antimalarial drug. Hence, in this paper, an image segmentation procedure is attempted to enhance the visibility of the Plasmodium sp. in microscopic blood smear images. In this paper, two RGB blood smear images of Plasmodium ovale (300 × 300) are considered and segmented using Otsu and heuristic algorithms, such as PSO, DPSO, and FODPSO available in the literature. During the segmentation procedure, maximization of a multiple objective function is adopted to guide the heuristic algorithm-based exploration. The performances of considered algorithms are analyzed using the popular image parameters, such as Otsu’s function, SSIM, RMSE, and PSNR. This study shows that FODPSO offers improved segmentation result compared to PSO and DPSO algorithms. The similar procedure can be used to identify other Plasmodium sp. using the microscopic blood smear images.

Solving Multi-level Image Thresholding Problem—An Analysis with Cuckoo Search Algorithm

In recent years, heuristic algorithms are extensively employed to offer optimal solutions for a class of engineering optimization problems. In this paper, Otsu based bi-level and multi-level image segmentation problem is addressed using Cuckoo Search (CS) algorithm. Optimal thresholds for the gray scale images are attained by analyzing histogram of the image. Maximization of Otsu’s between class variance function is chosen as the objective function. In the proposed work, CS algorithm with various search methodologies, such as Levy Flight (LF), Brownian Distribution (BD), and Chaotic search are analyzed. The proposed work is demonstrated by considering five grey scale benchmark (512 × 512) images. The performance assessment between CS algorithms are carried using established image parameters such as objective function, Root Mean Squared Error (RMSE), Peak to Signal Ratio (PSNR), and Structural Similarity Index Matrix (SSIM). The result shows that BD and chaotic CS provide better objective function, PSNR and SSIM, whereas LF based CS offers faster convergence.

Brownian Distribution Guided Bacterial Foraging Algorithm for Controller Design Problem

Bacterial Foraging Optimization (BFO) algorithm is widely adopted to solve a variety of engineering optimization tasks. In this paper, the Brownian Distribution (BD) strategy guided BFO algorithm is proposed. During the optimization exploration, BD monitors and controls the chemotaxis operation of the BFO algorithm inorder to enhance the search speed and optimization accuracy. In the proposed algorithm, after undergoing a chemotaxis step, each bacterium gets mutated by a BD operator. In the proposed work, this algorithm is employed to design the PID controller for an AVR system and unstable reactor models. The success of the proposed method has been confirmed through a comparative analysis with PSO, BFO, adaptive BFO and PSO + BFO based hybrid methods existing in the literature. The result shows that, for unstable reactor models, the BD guided BFO algorithm provides better optimization accuracy compared to other algorithms considered in this study.

An efficient clustering technique and analysis of infrared thermograms

This work proposes an efficient clustering technique for the localization of normal and abnormal tissues using the thermal data obtained from Digital Infrared Thermal Imaging. 10 normal and abnormal raw thermograms are preprocessed and by using K-means clustering, the heat patterns of the thermograms are clustered into various objects using the Euclidean distance metric. Further, breast thermograms are analysed, extracting the region of abnormality by utilizing the fuzzy nature of these thermograms. Features extracted from the simulations conducted on breast thermograms are compared and a distinctive variation is observed. These features can be used efficiently to identify normal and abnormal tissues.

Firefly Algorithm-Assisted Segmentation of Brain Regions Using Tsallis Entropy and Markov Random Field

In recent years, segmentation of medical images attracted the research community because of its significance in medical discipline. In this paper, firefly algorithm and Tsallis entropy-based approach is initially considered to threshold the standard brain MRI dataset. Later, the brain regions, such as white matter (WM), gray matter (GM), and cerebrospinal fluid (CF), are segmented using the Markov random field (MRF) approach. The proposed work is implemented using 256 × 256 sized benchmark MRI data, of subjects CHIMIC, JANPRZ, and JATKAM. Performance of the proposed approach is validated using a numerical metric that estimates the silhouette index of the estimated clusters. The proposed approach is also tested on other brain MRI dataset available in the literature and obtained better result in the segmentation of WM, GM, and CF. The simulation results in this study confirms that the proposed method offers an average enhancement of cluster classification by 4.44% in terms of silhouette index.

Preliminary big data analytics of hepatitis disease by random forest and SVM using r-tool

In the growing era of technology, concentration is on the analysis of large amount of structured and unstructured data. The processing applications are inadequate to deal with these data are termed as BigData since in large amounts. In this work, an initial stage for analysing medical informatics using R-studio by R programming is attempted by two algorithms. The biomedical data is used because they are concerned with the real time usage and is an open access journal aiming to facilitate the presentation, validation, use, and re-use of datasets, and can be modifiable with focus on publishing biomedical datasets that can serve as a source for simulation and computational modelling of diseases and biological processes. Random forest technique and support vector machine (SVM) techniques are used to derive features from the database and are able to differentiate various disease supports. The aim of this paper is to provide a comparison between the various techniques that are involved in the field of sorting the data and analysing them in large numbers. For this the process of data mining is used. Data mining is the process of extracting valuable information from a large set of databases. The latter technique produces more appropriate results that has less deviation from the reference taken from the hepatitis profile. By this method one can get the lead vision of the results that are produced by medical science. Therefore the SVM technique can be implemented practically in the medical field.