Fig. 1. A simple example.

Fig. 2. Incompatibilities association graph of the example in Fig. 1(a): dashed lines represent rejected arcs.

Fig. 3. The association graph of the example in Fig. 1(a) after merging regions.

Fig. 4. Synthetic stereo image pair, corridor (without noise).

Fig. 5. Segmentation by region merging of the corridor stereo pair and merged regions resulting from the algorithm.

Fig. 6. Ground truth disparity map and results from our algorithm.

Fig. 7. Segmentation by region clustering of the corridor stereo pair and merged regions resulting from the algorithm.

Fig. 8. Disparity map resulting from our algorithm.

Fig. 9. Results for a stereo pair of real images, the parking meter example [28], segmented by the region merging method.

Fig. 10. Results for a stereo pair of real images, the textured lamp example, segmented by the region merging method.

Fig. 11. Results for a stereo pair of real images, the lab example, segmented by the region merging method.

Fig. 12. Disparity maps and matching rates of the corridor example using the optimal and suboptimal algorithms (T_s=0.80).

Fig. 13. Disparity maps and matching rate (T_s=0.40).

Fig. 14. Noise effect in the matching rate.

Fig. 15. Relation between graph size and matching rate at different noise levels.

Fig. 16. Growth of execution time with respect to the number of nodes considered for finding the best maximal clique.

Table 1. Percentage of matching for the corridor pair (without noise)

Table 2. Percentage of matching for the corridor pair (without noise), using the suboptimal algorithm with K=30