A multi-scale gradient approach for color-based morphological segmentation

The watershed transform is known to be a powerful tool for morphological image segmentation. Usually, the watershed starts with the gradient of the image to be segmented. The watershed lines partition the gradient image into different catchment basins, which correspond to homogeneous regions. Consequently, the watershed result is significantly influenced by the morphological gradient chosen for the computation. In this paper, we propose a morphological approach to compute the multi-scale gradient applied to color images. The main property of this technique, established on color morphology, is that it does not split the color channels in contrast to other methods in the literature. The experiments show that the suggested technique enhances the segmentation results generating more precise watershed lines.