A New Method to Optimize Initial Cluster Number and Then Apply K-Means Algorithm to Perform Image Segmentation
Clustering is described as the task of merely grouping a set of data points, which in our case are represented by image pixels, with the application of one or more algorithms, into super-pixels. The k-means clustering algorithm is used for many day-to-day practical applications where the quality of the obtained clusters depends heavily on the initial selection of centroids. In literature, several methods have already been proposed for the improvement in the performance of the k-means clustering algorithm. This paper too proposes a method for optimizing the initial cluster number, thereby making the algorithm more effective and reducing the iterations performed, in order to get better clusters with reduced complexity. Image segmentation is a pre-processing step that plays a major role, especially in areas of object recognition, computer vision image analysis and tracking. Hence the final output results in the partial image segmentation of the input image. Hence, partial image segmentation is achieved where the image is segmented into different regions representing similar attributes like colour or intensity.
- There are currently no refbacks.