Color Image Segmentation in a Novel Dynamic Color Space

This paper presents a dynamic color model based on the electromagnetic wave theory to characterize the color feature of pixels, which simulates the continuous variety of visible light. In practice, the dynamic model can be discretized, and then we use a vector to represent the color feature of pixels in a K-dimension space, which is named the dynamic color space. To verify the dynamic space performance, an adopted segmentation method based on clustering techniques is proposed, which can be used to partition a color image in RGB space or in our dynamic color space. Experiments show that the segmentation results in our dynamic space is better than that in RGB space. Furthermore, the authors believe that the dynamic color space can be used to texture analysis or other image processing.