Abstract
In this paper, a novel approach of K-Region based Clustering image segmentation algorithm has been proposed. The proposed algorithm divides an image of size N × N into K number of regions. The K and N are multiples of 2. The value of K must be less than N. Authors divided the image into 4, 16, 64, 256, 1024, 4096 and 16384 regions, based on the value of K. The adjacent pixels having similar intensity value in each region are grouped into same clusters. Further, the clusters of similar values in each adjacent region are grouped together to form the bigger clusters. The different segmented images have been obtained based on the K number of regions. The four parameters, namely, Probabilistic Rand Index (PRI), Variation of Information (VOI), Global Consistency Error (GCE) and Boundary Displacement Error (BDE) have been used to evaluate the performance of the proposed algorithm. The performance of proposed algorithm was evaluated using 100 images taken from Berkeley image database. The time-complexity of the proposed algorithm has also been calculated. The comparative analysis of proposed algorithm was made with existing image segmentation algorithm, namely, K-mean clustering and Region-growing algorithm. Significant results were obtained in case of proposed algorithm when\the PRI, VOI, GCE and BDE values were compared with those of existing algorithms. MATLAB 7.4 has been used to implement the proposed algorithm.
Highlights
Image segmentation is the process of identifying and delineating objects in images
The segmentation parameters Probabilistic Rand Index (PRI), Global Consistency Error (GCE), Variation of Information (VOI) and Boundary Displacement Error (BDE) are used to evaluate the performance of the proposed algorithm
The VOI and BDE value of 16384 region-segmented image is low as compared to other segmented images
Summary
Image segmentation is the process of identifying and delineating objects in images. Image segmentation may consist of two related processes—recognition and delineation. The performance of Region-growing algorithm depends on the number of seed points in the image. The performance of K-mean clustering algorithm highly depends on the number of clusters in the image. Clusters in each adjacent region are merged together to obtainthe larger clusters In this way, the segmented image was obtained. The authors have taken 100 images from Berkeley image database[9] to evaluate and perform the comparative analysis of proposed, K-mean clustering and Region-growing algorithm. This paper is presented in different sections: Section 2 proposes the K-Region based clustering algorithm and has shown the working for image of size 4 × 4.
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call