Artifact

Oscar Danielsson
CSE455 Computer Vision
Project 2 Artifact

· Test sequence

· First sequence taken with Kaidan head

· Second sequence taken with Kaidan head

· Hand held sequence

Test sequence

[full resolution jpeg] [batch file for stitching]

First sequence taken with Kaidan head.

[full resolution jpeg] [panoramic viewer] [batch file for stitching]

I took the first sequence from the stairs between the second and third floors in Paul G. Allen building. In total the sequence consists of 18 images (not counting the first image twice).

Second sequence taken with Kaidan head.

[full resolution jpeg] [panoramic viewer] [batch file for stitching]

The second sequence is shot from the same place as the first, but at night. This sequence was mainly taken for comparison to the handheld sequence.

Handheld sequence.

[full resolution jpeg] [panoramic viewer] [batch file for stitching]

The handheld sequence was shot under the same conditions as the second Kaidan head sequence. Obviously the registration was far from perfect in this case. One possible reason for this is that I forgot to use the panorama tool in the camera. Because of this I probably did not keep the axis of camera rotation steady. This is a problem since the Lucas-Kanade algorithm assumes pure translational motion on the face of the projection cylinder, which corresponds to having a fixed axis of rotation. Furthermore I took 22 images instead of 18 for this sequence, so at some places I probably had 3 overlapping images.