Comparative Study of Color Image Segmentation by the Seeded Region Growing Algorithm

The choice of color representation can have distinguishable perceptual differences in the subject image which raises the following question: To what extent color representation can affect image processing results? In this paper, we study the effect of the RGB and HSV color representations on the segmentation result of the famous seeded region growing (SRG)algorithm. The implemented method involves three steps: 1) The automated seed selection, based on both color and space features,2) The region growing, based on the neighborhood similarity measured by the Euclidean distance, and finally,3) the region merging phase, introduced to overcome the over-segmentation issue and improve the results' accuracy. We used three metrics from the literature to evaluate the performances of our algorithm on both color spaces. The segmentation results were compared by combining the performance measures taken from a sample of images from the Berkeley dataset. The algorithm showcased more accurate results and consumed less execution time in the HSV color space compared to the RGB one.