Abstract: : Image segmentation of noisy image has nowadays gained popularity in the field of computer vision. This paper presents comparison between two density based clustering algorithms DBSCAN and Mean Shift. Considering an image as a dataset of pixels we firstly remove salt and pepper noise from an image using median filtering technique followed by applying DBSCAN algorithm to cluster it. Next, the implementation of Mean Shift algorithm is seen followed by the comparison of both the outputs. Density based clustering algorithms are used to find spatial connectivity and colour similarity of the pixels, which is used to discover clusters of arbitrary shape leading to the partitioning of pixels and further isolating the noise points. Experimental results using proposed method demonstrate encouraging performance.

Keywords: Image segmentation, Cluster, DBSCAN, Median Filter, Salt and Pepper noise.