Fig. 1. Left: the custom 3D range finder mounted on top of the mobile robot Kurt3D. Right: an office scene imaged with the 3D scanner in remission value mode, medium resolution (361 × 211 pixels).

Fig. 2. Visualized laser data. Left, scene from camera image; middle, visualized depth data; right, visualized remission data. Depending on the sensor, the presented images have slightly different extensions, the laser scanner getting a wider angle than the camera in all directions.

Fig. 3. The BILAS. The images from the two laser modes “depth” and “reflectance” are computed independently. Saliencies according to intensity and orientations are determined and fused into a mode-specific saliency map. After combining both of those maps, the focus of attention is directed to the most salient region.

Fig. 4. Algorithm of the BILAS.

Fig. 5. Top: the white pop-out is not detected by the NVT (left) but by BILAS (right). Only separating the on-center–off-surround difference from the off-center–on-surround difference enables the pop-out. Bottom: two intensity maps of a breakfast table scene, computed by the NVT (left) and by BILAS (right). The square-based structure in the left image resulting from taking the difference between two scales can be seen clearly, the right image shows a much more accurate solution.

Fig. 6. The first two foci of attention computed by BILAS on laser scanner data. Left: the scene in a camera image. Right: foci on the combination of range and reflection data.

Fig. 7. The foci in laser data show some advantages of the depth mode. In reading order: depth image, reflection image, combined image, camera image. The rubbish bin is salient only in the range data. Here, the stronger influence of the depth image causes the first focus to point to the rubbish bin in the combined image, too.

Fig. 8. The foci in laser data show some advantages of the depth mode. In reading order: depth image, reflection image, combined image, camera image. The open door is salient only in the range data.

Fig. 9. The foci in laser data show some advantages of the reflection mode. In reading order: depth image, reflection image, combined image, and camera image. The handicapped person sign is salient only in the reflection data.

Fig. 10. The foci in laser data show some advantages of the reflection mode. In reading order: depth image, reflection image, combined image, and camera image. All of the four cars are among the first six focus regions in the reflection data.

Fig. 11. Foci showing the same region in camera and in laser data. Left: a camera image with a focus on a red traffic sign generated by the NVT. Right: a laser image, combined from depth and reflection data, with a focus generated by BILAS. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this paper.)

Fig. 12. The foci show some advantages of camera images: the red car and the red telephone box are only focused in the camera images (left), but not in laser data (right). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this paper.)

Fig. 13. The foci show some advantages of the laser data: The traffic sign (ahead) and the person (below) are only focused in the laser data (right), but not in camera images (left). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this paper.)

Fig. 14. Salient regions detected by the attention system serve as starting points for a classifier. Top row: the first respectively the first 5 foci of attention computed on depth and reflection data. Bottom row: recognized objects in the focus regions.

Fig. 15. Salient regions detected by the attention system serve as starting points for a classifier. Top row: the first respectively the first 5 foci of attention computed on depth and reflection data. Bottom row: recognized objects in the focus regions.