Modified K-means Clustering Research Articles

Context Quantization Based on the Modified K-Means Clustering

The context quantization forsource based on the modified K-means clustering algorithm is present in this paper. In this algorithm, the adaptive complementary relative entropy between two conditional probability distributions, which is used as the distance measure for K-means instead, is formulated to describe the similarity of these two probability distributions. The rules of the initialized centers chosen for K-means are also discussed. The proposed algorithm will traverse all possible number of the classes to search the optimal one which is corresponding to the shortest adaptive code length. Then the optimal context quantizer is achieved rapidly and the adaptive code length is minimized at the same time. Simulations indicate that the proposed algorithm produces better coding result than the result of other algorithm.

An intelligent content-based image retrieval system for clinical decision support in brain tumor diagnosis

Accurate diagnosis is crucial for successful treatment of the brain tumor. Accordingly in this paper, we propose an intelligent content-based image retrieval (CBIR) system which retrieves similar pathology bearing magnetic resonance (MR) images of the brain from a medical database to assist the radiologist in the diagnosis of the brain tumor. A single feature vector will not perform well for finding similar images in the medical domain as images within the same disease class differ by severity, density and other such factors. To handle this problem, the proposed CBIR system uses a two-step approach to retrieve similar MR images. The first step classifies the query image as benign or malignant using the features that discriminate the classes. The second step then retrieves the most similar images within the predicted class using the features that distinguish the subclasses. In order to provide faster image retrieval, we propose an indexing method called clustering with principal component analysis (PCA) and KD-tree which groups subclass features into clusters using modified K-means clustering and separately reduces the dimensionality of each cluster using PCA. The reduced feature set is then indexed using a KD-tree. The proposed CBIR system is also made robust against misalignment that occurs during MR image acquisition. Experiments were carried out on a database consisting of 820 MR images of the brain tumor. The experimental results demonstrate the effectiveness of the proposed system and show the viability of clinical application.

Application of Modified K-Means Clustering Algorithm for Satellite Image Segmentation based on Color Information

The Image segmentation is the process of clustering or partitioning the image into number of sub images based on any one characteristics of the image such as colour, intensity or texture. The segmentation is the one of the middle level important process in the image analysis. The number of segmentation algorithms has been developed for various applications. In the field of satellite image processing, the segmentation is one of the vital step for gathering huge amount of information from the satellite images. The basic k-means clustering algorithm is simple and fast. The main problem associated with this clustering is not producing the same result for every run because the resulting clusters depends on the initial random assignments. In this paper, a modified k-means clustering algorithm is proposed for the effective segmentation of the satellite images. This proposed method always produces the same result for every run. The experimental results proved that the improved k-means algorithm is an efficient and effective method for the satellite image segmentation for the exact and accurate segmentation of satellite images. Keywords: image segmentation; satellite image; kmeans clustering; centroid; fuzzy logic

A Novel High-Quality Community Detection Algorithm Based On Modified K-Means Clustering

Some existing community detection algorithms have the defect that they need extra parameters (e.g. the number of communities) as input, while other algorithms needn’t to know prior information but have parameters notoriously difficult to adjust. In this paper, we present a new community detection algorithm based on modified k-means clustering method to overcome these problems. In our algorithm, we first present a new initialization method that can find initial cores no matter the number of communities is known or not and each core is regarded as a initial community. We also propose a novel two-stage method to assign nodes to corresponding communities, and then recalculate the core of each community. By repeating steps above until convergence, we get a division of a network. A series of experimental results demonstrate the performance of our proposed algorithm.

Image Fusion Base on K-Means Clustering and Contourlet Transform

The robustness of K-means clustering is poor in non-spherical distribution data, in order to improve the universal ability of clustering algorithms, the cross-entropy distance measure was used to replace the Euclidean distance measure . Contour let transform, not only has characteristics of multi-resolution, locality and critical sampling which wavelet has, but also has the characteristics of multiple decomposition directions and anisotropy which wavelets lack. So we combine the modified K-means clustering and Contour let transform to apply for image fusion. Experimental results show that this method is feasible.