Debasis Chaudhuri

A Novel Objective Function Based Clustering with Optimal Number of Clusters

The clustering is one of the fundamental steps in the field of data mining. This is widely used in various real-life applications for realizing potential information from the datasets. However, this particular procedure is intensive in term of computation. So, the clustering procedure is preferred with the optimal number of clusters. In this paper, the authors have suggested a novel objective function which is more effective in clustering with the optimal number of clusters detected through cluster validity indices. The proposed objective function based clustering approach is implemented on different types of images and the outcomes depict effective performance in terms of cluster quality based on segmentation entropy, and cluster partitioning time. The results are comparable with other related works and the satisfactory outcome is attained.

Seed Point Selection Algorithm in Clustering of Image Data

Massive amount of data are being collected in almost all sectors of life due to recent technological advancements. Various data mining tools including clustering is often applied on huge data sets in order to extract hidden and previously unknown information which can be helpful in future decision-making processes. Clustering is an unsupervised technique of data points which is separated into homogeneous groups. Seed point is an important feature of a clustering technique, which is called the core of the cluster and the performance of seed-based clustering technique depends on the choice of initial cluster center. The initial seed point selection is a challenging job due to formation of better cluster partition with rapidly convergence criteria. In the present research we have proposed the seed point selection algorithm applied on image data by taking the RGB features of color image as well as 2D data based on the maximization of Shannon’s entropy with distance restriction criteria. Our seed point selection algorithm converges in a minimum number of steps for the formation of better clusters. We have applied our algorithm in different image data as well as discrete data and the results appear to be satisfactory. Also we have compared the result with other seed selection methods applied through K-Means algorithm for the comparative study of number of iterations and CPU time with the other clustering technique.

A new image segmentation technique using bi-entropy function minimization

Image segmentation, the splitting of a multispectral and panchromatic image into groups of homogeneous pixels based on the region of interest(ROI), is a universal step for many advanced image processing and object recognition. Image segmentation essentially affects the overall performance of any automated image analysis system due to utmost importance of its quality. Image segmentation can be performed by recursively splitting the whole image or by merging together a large number of minute regions until a specified condition is satisfied. Thresholding is an old, simple and important method in gray scale image segmentation. In this paper, we have used Shannon’s entropy and proposed a new multilevel thresholding image segmentation method based on minimization of bi-entropy function. A smoothing technique based on weight value of the pixel within a w × w moving window is introduced to make the splitting result continuous and qualitative. The proposed algorithm takes full account of the spatial information and the gray information to decrease the computing quantity. Standard medical images, texture images, and remote sensing images are segmented in the experiment and compared with other related segmentation methods with different measures. Experimental results show that the proposed method can quickly converge with high computational efficiency.