Volume 21, Issue 3

Abstract. Textline segmentation in handwritten documents is of real challenge and interesting in the field of document image processing. In this paper, we propose a handwritten textline segmentation scheme based on the concept of linked list. The proposed method consists of three stages, namely, preprocessing, linked list and mathematical morphology. The concept of the linked list approach is used to build a textline sequence and mathematical morphology is used to obtain the line separator. We experimentally evaluated our proposed method on a document containing handwritten Kannada script. The results are compared with recent methods and show encouraging results.

Abstract. This paper illustrates the applications of various ensemble methods for enhanced classification accuracy. The case in point is the Pima Indian Diabetic Dataset (PIDD). The computational model comprises of two stages. In the first stage, k -means clustering is employed to identify and eliminate wrongly classified instances. In the second stage, a fine tuning in the classification was effected. To do this, ensemble methods such as AdaBoost, bagging, dagging, stacking, decorate, rotation forest, random subspace, MultiBoost and grading were invoked along with five chosen base classifiers, namely support vector machine (SVM), radial basis function network (RBF), decision tree J48, naïve Bayes and Bayesian network. The k -fold cross validation technique is adopted. Computational experiments with the proposed method showed an improvement of 16.14% to 22.49% in the classification accuracy compared to literature survey. Among the ensemble methods tried, MultiBoost ensemble with SVM classifier and grading ensemble with naïve Bayes showed the best performance followed by MultiBoost, stacking and grading ensemble with Bayesian classifier, rotation forest ensemble with RBF and grading and rotation forest ensemble with J48. This investigation conclusively proves the significance of cascading k -means clustering with ensemble methods in the enhanced accuracy in categorization of diabetic dataset.

Abstract. A paraconsistent logic is a logical system that attempts to deal with contradictions in a discriminating way. In an earlier paper [Notre Dame J. Form. Log. 49 (2008), 401–424], we developed the systems of weakening of intuitionistic negation logic, called and , in the spirit of da Costa's approach by preserving, differently from da Costa, the fundamental properties of negation: antitonicity, inversion and additivity for distributive lattices. Taking into account these results, we make some observations on the modified systems of and , and their paraconsistent properties.

Abstract. The problem of reconstructing digital images from degraded measurements is regarded as a problem of importance in various fields of engineering and imaging science. The main goal of denoising is to restore a noisy image to produce a visually high quality image. In this paper, we propose a novel transform domain technique that uses multispinning for image denoising. The proposed method uses multiple cyclic shifted versions of an image, where each of them would capture more detail information during decomposition. Discrete wavelet transform (DWT) and contourlet transform (CT) in association with multispinning is used. The results are compared with traditional transform (soft thresholding) and spatial domain techniques. The visual and quantitative evaluation suggests that the proposed method yields better results.

Abstract. Many large companies transact from multiple branches. It results in generating multiple databases, since local transactions are stored locally. The number of multi-branch companies as well as the number of branches of a multi-branch company is increasing over time. Thus, it is important to study data mining on multiple databases. Global exceptional patterns describe interesting individuality of few branches. Therefore, it is interesting to identify such patterns. In this paper, we propose type I and type II global exceptional frequent itemsets in multiple databases by extending the notion of global exceptional frequent itemset. Also, we propose the notion of exceptional sources for a type II global exceptional frequent itemset. We propose type I and type II global exceptional association rules in multiple databases by extending the notion of global exceptional association rule. We propose an algorithm for synthesizing type II global exceptional frequent itemsets. Experimental results are presented on both real and synthetic databases. We compare the proposed algorithm with the existing algorithm theoretically as well as experimentally. The experimental results show that the proposed algorithm is effective and promising.