Project 4- Panoramas
Sonja Duric
Original Pictures
Below are five pairs of pictures to be used for creating mosaics.
A view of Gates from in front of Smith Hall. | Cubicles in the Engineering and Science Library, Fourth floor Wean Hall. | A view down a table in Hunt Library. | A close-up view of a the Maggie Murph Cafe sign in Hunt Library. | The view of my front parking lot from my bedroom window, off campus. |
Points Chosen by Hand
These points were chosen by hand using the cpselect tool. Below are screenshots.
Harris Point Overlays for the Original Images
Below are all five pairs of images with Harris points overlayed on the images. These are the points that remain after non-maxima suppression using an 11x11 window.
Inliers and Homography
Here are the largest sets of inliers, as computed by 10,000 iterations of RANSAC, for each set of images and the respective homographys calculated on these points.
12 inliers | 2 inliers | 881 inliers | 774 inliers | 65 inliers | |||||
[0.1151, -0.1672, 0.4356 | [0.0000, 0, 0 | [0.8083, -0.0101, -0.0002 | [0.8849, -0.1079, 0.2841, | [0.0264,-0.1138,-0.0017, | |||||
0.0722, -0.2590, 0.3895 | 0, 0.0000, 0 | -0.0914, 0.8817, -0.0001 | 0.0089, 0.8660, 0.7857, | 0.1288, 0.3968, 0.0003 | |||||
-0.0001, -0.0026, 1.0000] | 0.0000, -0.0016, 1.0000] | -0.1916, 1.1414, 1.0000] | -0.0000, -0.0002, 1.0000] | 0.1803, 0.3979, 1.0000 |
Conclusions
It is clear from the sparse inliers in the first and second sets, that it is not possible automatically form a mosaic from these images. Not only are the inliers sparse but there is little to no correlation between these points upon inspection between the images. However, the last 3 pairs of images are much more dense with inliers and these points appear to be correlated upon inspection so it is possible to use these points/images to create a mosaic of the images.