Colour Based Segmentation of a Landsat Image Using K-Means Clustering Algorithm
Image segmentation is one of the typical vintage subjects in image processing and it acts as a bulls-eye of the image processing technique. By definition, image segmentation means identifying the similar regions in the image; or in other words, identifying the homogenous pixels in the image and grouping all these pixels based on the homogeneity condition considered. This homogeneity condition can be like, colour, texture, size, compactness etc. It is very vital to categorize the segmentation techniques to interrelate with the growing needs in image segmentation. Depending upon the segmentation accuracy, the success of image analysis varies. The demanding task is to segment the image into segments based on the colour homogeneity condition considered. One of the most popular unsupervised clustering techniques is the K-Means clustering algorithm which is used for the segregation of the image into multiple regions based on the colour image property. This paper is focused on the K-Means clustering technique based on the cluster index values. K value can be manually adjusted to meet with the needs of each image. This K value is then passed to the K-Means clustering algorithm followed by K nearest neighbour (k-NN) classifier for image segregation. This proposed approach produced very good results with high resolution images as well as with low resolution (30 m) satellite (Landsat) images. The algorithms are implemented using MATLAB.
Keywords: K-Means clustering, k nearest neighbour, segmentation, classification, Landsat
Cite this Article
Rohith John, Ramesh H. Colour Based Segmentation of a Landsat Image Using K-Means Clustering Algorithm. Journal of Image Processing & Pattern Recognition Progress. 2017; 4(3): 31–38p.
- There are currently no refbacks.