Segmentation of color images using mean curvature flow and parametric curves (Q830048): Difference between revisions

In the paper, the authors study the segmentation problem of colour images. The authors focus on incorporating the CIE94 colour distance function and the enhancement of the splitting algorithm. In the proposed method, the colour image is first transformed into a grey-scale image using a colour difference function. For this aim, the CIE94 version of the colour difference function is used, but before using this function, the authors convert the image from the RGB space to the Lab colour space, which is required by the used colour function. Next, the parametric mean curvature flow equation is used to segment the grey-scale images. Because the parametric approach does not handle splitting or merging curves when they touch themselves, the authors propose a numerical algorithm to cover this disadvantage for two open curves which touch themselves during the evolution. To show the effectiveness of the proposed approach, the authors show several examples of segmentation of artificial colour images and on real photos as well.