Color Image Segmentation Based on Threshold Selection

Color image segmentation can be seen as an extension of the grayscale image segmentation. In order to provide basic data for image recognition and image segmentation, this paper studies the image under consideration based on the threshold selection method, Otsu (named after Nobuyuki Otsu). Firstly, pre-processing operation is carried out on the unprocessed original image. Then, we select The L*a*b* color space as the optimum color space for image segmentation, the color image of RGB color space is transformed to L*a*b* color space. Next, the channels of color space are separated and then a single channel is selected after which two-dimension Otsu segmentation is carried out based on the selected channel. This proposed method automatically performs histogram-shape based image thresholding. Experiments show that Otsu method will lead to a correct threshold value and yield an ideal result in segmentation with small computation time.