Color Image Segmentation Algorithm based on RGB channels

This paper presents a new region growing based color image segmentation technique with the use of RGB channels. The seed pixel selection and region growing phases are based on 8-neighbour connectivity similarity measurement. The region growing process based on intensity measurement between the seed pixel and 8-neighbour pixels of seed pixel. Region growing is applied to Red channel; Green channel and Blue channel separately, then combined labeled matrix is created to obtain segmented image. The threshold value (mu) is used to stop region growing process. Random pixel elimination method merges small regions to large neighbor regions in segmented image. The proposed segmentation technique is performed and analyzed with set of images obtained from Berkley Image segmentation dataset. The result of proposed technique is presented with the help of image processing tool.