Kluwer Academic Publishers Color Segmentation Algorithm Using an Hls Transformation

A new segmentation procedure for color images is detailed in this article. The segmentation still ends with the watershed function, but the whole series of pre-treatments applied to the color images is of interest. After a HLS transformation of the images, a color reduction is applied, followed by a median filtering. The gradient is then constructed from this resulting image. After this step, two different ways to obtain the boundaries are proposed: either a hierarchical segmentation or a simple watershed line followed by a merging of the bassins according to the color of their minima. AlI these steps are detailed in the following and the results obtained on test images are discussed.