next up previous
Next: Initialisation: Perceptual Edges Up: Colour Texture Segmentation by Previous: Related Work


Image Segmentation Strategy

As stated in our review work [5], the different integration strategies try to solve different problems that appear when simple approaches (region or boundary-based) are used separately. Hence, we consider that some of these strategies are perfectly complementary and it could be greatly attractive to fuse different strategies to perform the integration of region and boundary information. The fusion of several approaches will allow to tackle an important number of issues and to exploit at maximum the possibilities offered by each one. Hence, we propose an image segmentation method which combines the guidance of seed placement, the control of decision criterion and the boundary refinement approaches.

Figure 1: Scheme of the proposed colour texture segmentation strategy.
\includegraphics[height=5.8 cm]{images/strategy.eps}

Our approach uses the perceptual edges of the image to adequately place a set of seeds in order to initialise the active regions. The knowledge extracted on these regions allows to define the region information and to extract accurate boundary information. Then, as these regions grow, they compete for the pixels of the image by using a decision criterion which ensures the homogeneity inside the region and the presence of edges at its boundary. A scheme of the proposed strategy is shown in Figure 1. The inclusion of colour texture information into our initial segmentation proposal involves two major issues: 1) the extraction of perceptual edges, 2) the modelling of colour and texture of regions.



Subsections
next up previous
Next: Initialisation: Perceptual Edges Up: Colour Texture Segmentation by Previous: Related Work
Xavier Llado 2004-05-31