In this assignment, you need to finish the Automatic Panoramic Image Stitching process from scratch, excluding the first step. You also need to show the results of your own data in the report. Following is the steps:
Interest points detection & feature description by SIFT You can use OpenCV function in this part.
Feature matching by SIFT features
You can try ratio distance(last lecture) in this part:
It is also good for you to try different features, e.g. MSER
The result of this part need to be shown like
You can have better visualization, for example more colorful:)
RANSAC to find homography matrix H You need a function:
H = homomat(points_in_img1, points_in_img2)
In this part you have to write a RANSAC algorithm to find out the best homography matrix between two images.
*RANSAC hints:
sample S correspondences from the feature matching results
compute the homography matrix based on these sampled correspondences
check the number of inliers/outliers by a threshold
iterate for N times
get the best homography matrix with smallest number of outliers
Warp image to create panoramic image
You need to write a function warp(img1, img2, H) which warps one image onto the other using homography and provide examples.
*Warp hints:
In this part you are also encouraged to use a blending function to enhance the result.
l Important notice:
Everyone is supposed to upload their own zip file. You should follow the file name form: CV_HW3_GroupNumber_ID.zip (for example, CV_HW3_43_0210519.zip)
In zip file, it should contain your code and report. The report should be a pdf file and be written in English.
Please submit your homework before the deadline!!!
