Enhanced Algorithm of Superpixel Segmentation Using Simple Linear Iterative Clustering

Abstract

Image segmentation process represents the main stage for most computer vision systems. This paper presents an improved algorithm based on simple linear iterative clustering (SLIC) to reduce the number of used seeds for threshold estimation as well as the entire execution time of image segmentation. These is achieved by using split and merge stages for the location, number of seeds as well as other parameters of the seed points. The obtained results showed the possibility of using various threshold levels instead of a single one which represented a challenge due to the estimation complexity. The independent of the threshold-level estimation can contribute significantly in improving performance of the overall image segmentation process).