CSE 473/573 – Computer Vision and Image Processing
1 Image Stitching (7 points)
The goal of this task is to stitch two images together to construct a panorama image. Please follow
the following steps to complete the task. (1) find keypoints (points of interest) in both images, e.g.,
Harris detector or SIFT point detector. (2) use SIFT feature descriptors to extract features for these
keypoints. (3) match the keypoints between two images by comparing their feature similarities. (4)
compute the homography matrix using RANSAC algorithm. (5) use the homography matrix to
stitch the two images. Figure 3 shows an exemplar result, while Figure 4 is the failed case. Note
Figure 1 and Figure 2 are just examples, and the two images you need to stitch are provided in the
Guidelines You can use APIs provided by numpy and opencv to complete the task, except for
APIs that have “stitch” or “Stitch”, “match” or “Match” in their names, e.g., “cv2.BFMatcher()”
and “cv2.Stitcher.create()”. You will be provided with a “task1.py” file and two original images.
Please strictly follow the format in the “task1.py” and do not modify the code format provided to
you. You need to provide the output panorama image “task1 result.jpg” after image stitching in
the results folder.
Figure 1: Left Image Figure 2: Right Image
CSE 473/573 Homework 1 25 Nov. 2019
Figure 3: Successful example
Figure 4: Failed example
2 Hough transform (8 points)
You are given an image ‘hough.jpg’ shown in figure 5, please design and implement an algorithm
to detect lines and circles.
• Utilizing edge detection and Hough transformation to detect all vertical lines (red lines), save
the result image name as ‘red lines.jpg’ and the line parameters in ‘red lines.txt’. (2 points)
• Detect all the diagonal lines (blue lines), save the result image as ‘blue lines.jpg’ and the line
parameters in ‘blue lines.txt’ (3 points)
• Detect all the coins in the image, save the result image as ‘coins.jpg’ and the circle parameters
in ‘coins.txt’. ( 3 points)
• For all the red and blue lines detected, you should report the θ and ρ parameters in ‘red lines.txt’,
‘blue lines.txt’ files respectively. For each of the detected line report θ and ρ considering line
is represented by equation xcosθ + ysinθ = ρ. (Please refer slide 60 in Image Analysis slides
in piazza). Each line in the .txt file will contain parameters of a single line in [θ, ρ] format.
You can assume that he resolution of ρ is 1 pixel and resolution of θ is 1 degree.
• For all the detected circles, please report c1, c2, c3 in the ‘coins.txt’ file, considering that a
circle is represented by equation (x − c1)
2 + (x − c2)
2 = c3
. (Please refer slide 62 in Image
Analysis slides in piazza). Each line in the txt file will contain parameters of a single circle in
[c1, c2, c3] format.
• File naming conventions and the format of *.txt files needs to be strictly followed.
本网站支持淘宝 支付宝 微信支付 paypal等等交易。如果不放心可以用淘宝交易！
E-mail: [email protected] 微信:itcsdx