Calculate moving average of 3x3 homography matrices -- 2
Budget: $250 – $750 USD
The total number of 3x3 homography matrices is around 160K.
Each homography matrix projects the image from camera- (cam_view_1.jpg) to top-view (top_view_1.jpg).
Note that the reference-points (red-dots) may vary from one image to another and hence the scale of the homography matrix.
The homography matrix (H1) for cam_view_1 -> top_view_1 is:
-156399.0362570212, -812316.5807606786, 143090519.29495946
24449.701467687442, -982407.3062078588, 147784018.8125531
-11.330804255111957, -736.1835040833266, 28634.49043049858
The homography matrix (H2) for cam_view2 -> top_view_2 is:
-0.4057195728473708, -2.3341038499333258, 416.54208240668316
0.06659081277911126, -2.8283531710681795, 427.1502617756748
-3.2471167623300936e-05, -0.002160010952378696, 0.12255084544102246
Both matrices project on the same space (slight change in camera-rotation.
Note that the scale is vastly different.
Performing a SVD (H_i = U_i A_i V_i), shows that U_1 is similar to U_2 and V1 is similar to V2 whereas the eigenvalues, i.e., A_1 and A_2 are vastly different.
Note that a simple (H1 + H2)/2 does not suffice.
Tasks:
Rescale H_1 and H_2 such that the average can be used to warp both images from camera- to top-view.
In order to be elected you will need to send me the rescaled version of H1 and H2 as well as its average to proof that you are capable of doing is task.
If your bid is supposed to be considered use the word MATRIX (automatic bids are ignored).
Each homography matrix projects the image from camera- (cam_view_1.jpg) to top-view (top_view_1.jpg).
Note that the reference-points (red-dots) may vary from one image to another and hence the scale of the homography matrix.
The homography matrix (H1) for cam_view_1 -> top_view_1 is:
-156399.0362570212, -812316.5807606786, 143090519.29495946
24449.701467687442, -982407.3062078588, 147784018.8125531
-11.330804255111957, -736.1835040833266, 28634.49043049858
The homography matrix (H2) for cam_view2 -> top_view_2 is:
-0.4057195728473708, -2.3341038499333258, 416.54208240668316
0.06659081277911126, -2.8283531710681795, 427.1502617756748
-3.2471167623300936e-05, -0.002160010952378696, 0.12255084544102246
Both matrices project on the same space (slight change in camera-rotation.
Note that the scale is vastly different.
Performing a SVD (H_i = U_i A_i V_i), shows that U_1 is similar to U_2 and V1 is similar to V2 whereas the eigenvalues, i.e., A_1 and A_2 are vastly different.
Note that a simple (H1 + H2)/2 does not suffice.
Tasks:
Rescale H_1 and H_2 such that the average can be used to warp both images from camera- to top-view.
In order to be elected you will need to send me the rescaled version of H1 and H2 as well as its average to proof that you are capable of doing is task.
If your bid is supposed to be considered use the word MATRIX (automatic bids are ignored).