Search

The Accurate segmentation and classification takes place a major role in the medical image processing to detect and locate the abnormal tissue region. In this, the three different types of brain magnetic resonance imaging (MRI) image source such as Type-1, Type-2 and Fluid attenuated inversion recovery are combined by the image registration process to detect the clear region of the tumor tissue since, the region of interest identification in the single image data contains less key points to define it. In this paper, we implement the ensemble technique of image segmentation to segment the tumor region of the brain MRI image. For the segmentation process, the images are pre-processed by Laplacian cellular automata filtering method and segmented by ensemble of different clustering method such as K-means, fuzzy based clustering, self-organization map (SOM) and ensemble of Gaussian mixture model, K-means, SOM and their results are compared. This ensemble cluster label is consider as the segmented result and classify the abnormalities by using deep super learning method. The experimental results and the comparison charts defines the performance rate of proposed method comparing to the other state-of-art methods. The average accuracy for the proposed work is 98% in Ensemble 1 and 97% in Ensemble 2 methods for the BraTS brain image dataset.

A simple and effective recognition system for indoor scenes is presented. The proposed system has two phases, namely, creation of mandatory and desirable objects and an indoor scene recognition system (ISRS). In the first phase, a list of probable objects and their classification, such as mandatory and desirable objects, for any generic scene is created based on real-time indoor environment clubbed with human knowledge on standard datasets. In the second phase, the proposed system contains four stages. In the first stage, the proposed ISRS identifies and recognizes the objects of the given key frame based on a simplified version of CNN architecture of YOLO v3. In the second stage, the identified objects are divided into two sets of mandatory and desirable objects with a simple dictionary look-up. In the third stage, the objects are identified to belong to a probable scene, and this technique is called scene-object identification. In the final stage, a binary scene representation (BSR) is proposed for each probable scene, and a final scene recognition is obtained with a new scene-score, obtained after converting the binary BSR into a decimal number. The effect of the proposed ISRS has been experimented with standard datasets and measured in terms of standard metrics, besides comparison with existing schemes. The results are encouraging.

Distributed data mining technology has become an eloquent tool for recognizing patterns and designating ideas from large pools of data procured from different sections (multiple parties). The primary focus of this research is to formulate association rules that are accepted universally, restraining the information shared about each party. A privacy-preserving algorithm is projected to mine association rules from horizontally partitioned data through a new Inverse Frequent Item set Tree. This tree was formulated using inverse frequent items and propagated only with the collaboration parties from whom the data was to be merged. A central third party was utilized in order to mine the association rules with infrequent item sets. The proposed approach constitutes a guided assistance in preserving corporate privacy, furthered by experimenting (analyzing) with the large data sets from the UCI machine learning repository. The resultant output from the approach showed positivity of preserving data with a high degree of security in multiparty computation.

P2P networks are generally used for file sharing applications. In this paper, we argue that existing methods of building heterogeneous unstructured peer-to-peer (P2P) networks has joining process and the rebuilding process which does not make effective load balancing. Since the joining process is to use random walk to assist new incoming peers, but it does not decide whether the new peer can be used for processing the workload. Similarly the rebuilding process specifies how the nodes should react when they lose links, that is it leaves the network safely without affecting the existing performance, but the peer machine that is leaving the network does not allocate its work to the remaining peers. So we have proposed a simple protocol for building heterogeneous unstructured peer-to-peer (P2P) networks. Our new protocol consists of two parts - the allocating process and the reallocating process. The basic idea for the allocating process is to use random walk to assist new incoming peers in selecting their suitable neighbors in terms of capacity and connectivity to achieve load-balancing and also to decide whether that peer can be used for processing the workload of the network based on its capacity. The reallocating process specifies how the node allocates its work to the remaining peers when they leave the network. Our protocol provides a guideline to optimize a P2P network by achieving effective load balancing.