Computer Vision Homework 4 Feature Descriptors, Homographies & RANSAC

16720B: Computer Vision Homework 4
Feature Descriptors, Homographies & RANSAC

Total points: 165, EC: 40

Instructions/Hints
1. Integrity and collaboration: Students are encouraged to work in groups but
each student must submit their own work. Include the names of your collaborators
in your write up. Code should NOT be shared or copied. Please DO NOT use
external code unless permitted. Plagiarism is prohibited and may lead to failure
of this course.
2. Start early! This homework will take a long time to complete.
3. Questions: If you have any question, please look at piazza first and the FAQ page
for this homework.
4. All questions marked with a Q require a submission.
5. For the implementation part, please stick to the headers, variable names,
and file conventions provided.
6. Attempt to verify your implementation as you proceed: If you don’t verify
that your implementation is correct on toy examples, you will risk having a huge
mess when you put everything together.
7. Do not import external functions/packages other than the ones already
imported in the files: The current imported functions and packages are enough
for you to complete this assignment.
8. Submission: We have provided a script check file.py which will check if you
have all the files needed for submission. The submission is on Gradescope, you
will be submitting both your writeup and code zip file. The zip file,
<andrew-id.zip contains your code and any results files we ask you to save.
Note: You have to submit your writeup separately to Gradescope, and
1
include results in the writeup. Do not submit anything from the data/ folder
in your submission.
9. Assignments that do not follow this submission rule will be penalized up to 10%
of the total score.
10. Please make sure that the file paths that you use are relative and not absolute.
Introduction
In this homework, you will implement an interest point (keypoint) detector, a feature
descriptor and an image stitching tool based on feature matching and homography.
Interest point (keypoint) detectors find particularly salient points in an image. We
can then extract a feature descriptor that helps describe the region around each of the
interest points. SIFT, SURF and BRIEF are all examples of commonly used feature
descriptors. Once we have extracted the interest points, we can use feature descriptors to match them (find correspondences) between images to do interesting things like
panorama stitching or scene reconstruction.
We will implement the BRIEF feature descriptor in this homework. It has a compact
representation, is quick to compute, has a discriminative yet easily computed distance
metric, and is relatively simple to implement. This allows for real-time computation, as
you have seen in class. Most importantly, as you will see, it is also just as powerful as
more complex descriptors like SIFT for many cases.
After matching the features that we extract, we will explore the homography between
images based on the locations of the matched features. Specifically, we will look at the
planar homographies. Why is this useful? In many robotics applications, robots must
often deal with tabletops, ground, and walls among other flat planar surfaces. When two
cameras observe a plane, there exists a relationship between the captured images. This
relationship is defined by a 3 × 3 transformation matrix, called a planar homography.
A planar homography allows us to compute how a planar scene would look from a
second camera location, given only the first camera image. In fact, we can compute
how images of the planes will look from any camera at any location without knowing
any internal camera parameters and without actually taking the pictures, all using the
planar homography matrix.
1 Keypoint Detector (25 pts)
We will implement an interest point detector similar to SIFT. A good reference for its
implementation can be found in [4]. Keypoints are found by using the Difference of
Gaussian (DoG) detector. This detector finds points that are extrema in both scale and
space of a DoG pyramid. This is described in [1], an important paper in computer vision.
Here, we will implement a simplified version of the DoG detector described in Section 3
of [4].
2
Figure 1: Example Gaussian pyramid for model chickenbroth.jpg
Figure 2: Example DoG pyramid for model chickenbroth.jpg
HINT: All the functions to implement here are located in keypointDetect.py.
NOTE: The parameters to use for the following sections are:
σ0 = 1, k =
√
2, levels = [−1, 0, 1, 2, 3, 4], θc = 0.03 and θr = 12.
1.1 Gaussian Pyramid
In order to create a DoG pyramid, we will first need to create a Gaussian pyramid.
Gaussian pyramids are constructed by progressively applying a low pass Gaussian filter
to the input image. This function is already provided to you in keypointDetect.py.
GaussianPyramid = createGaussianPyramid(im, sigma0, k, levels)
The function takes as input an image which is going to be converted to grayscale
with intensity between 0 and 1 (hint: cv2.cvtColor(...)), the scale of the zeroth level
of the pyramid sigma0, the pyramid factor k, and a vector levels specifying the levels
of the pyramid to construct.
At level l in the pyramid, the image is smoothed by a Gaussian filter with σl = σ0k
l
.
The output GaussianPyramid is a R × C × L matrix, where R × C is the size of the
input image im and L is the number of levels. An example of a Gaussian pyramid
can be seen in Figure 1. You can visualize this pyramid with the provided function
displayPyramid(pyramid).
1.2 The DoG Pyramid (5 pts)
The DoG pyramid is obtained by subtracting successive levels of the Gaussian pyramid.
Specifically, we want to find:
Dl(x, y, σl) = (G(x, y, σl−1) − G(x, y, σl)) ∗ I(x, y) (1)
3
Figure 3.a: Without edge
suppression
Figure 3.b: With edge suppression
Figure 3: Interest Point (keypoint) Detection without and with edge suppression for
model chickenbroth.jpg
where G(x, y, σl) is the Gaussian filter used at level l and ∗ is the convolution operator.
Due to the distributive property of convolution, this simplifies to
Dl(x, y, σl) = G(x, y, σl−1) ∗ I(x, y) − G(x, y, σl) ∗ I(x, y) (2)
= GPl − GPl−1 (3)
where GPl denotes level l in the Gaussian pyramid.
Q 1.2: Write the following function to construct a Difference of Gaussian pyramid:
DoGPyramid, DoGLevels = createDoGPyramid(GaussianPyramid, levels)
The function should return DoGPyramid an R × C × (L − 1) matrix of the DoG
pyramid created by differencing the GaussianPyramid input. Note that you will have
one level less than the Gaussian Pyramid. DoGLevels is an (L−1) vector specifying the
corresponding levels of the DoG Pyramid (should be the last L−1 elements of levels).
An example of the DoG pyramid can be seen in Figure 2.
1.3 Edge Suppression (5 pts)
The Difference of Gaussian function responds strongly on corners and edges in addition
to blob-like objects. However, edges are not desirable for feature extraction as they are
not as distinctive and do not provide a substantially stable localization for keypoints.
Here, we will implement the edge removal method described in Section 4.1 of [4], which
is based on the principal curvature ratio in a local neighborhood of a point. The paper
4
presents the observation that edge points will have a “large principal curvature across
the edge but a small one in the perpendicular direction.”
Q 1.3: The following function is provided to you because of its similar nature of the
Harris corner implementation you’ve previously done:
PrincipalCurvature = computePrincipalCurvature(DoGPyramid)
The function takes in DoGPyramid generated in the previous section and returns PrincipalCurvature,
a matrix of the same size where each point contains the curvature ratio R for the corresponding point in the DoG pyramid:
R =
Tr(H)
2
Det(H)
=
(λmin + λmax)
2
λminλmax
(4)
where H is the Hessian of the Difference of Gaussian function (i.e. one level of the DoG
pyramid) computed by using pixel differences as mentioned in Section 4.1 of [4]. (hint:
use Sobel filter cv2.Sobel(...)).
H =
?
Dxx Dxy
Dyx Dyy?
(5)
This is similar in spirit to but different than the Harris corner detection matrix [3] you
saw in class. In your report, please explain the two methods’ similarity and dissimilarity.
We can see that R reaches its minimum when the two eigenvalues λmin and λmax
are equal, meaning that the curvature is the same in the two principal directions. Edge
points, in general, will have a principal curvature significantly larger in one direction
than the other. To remove edge points, we simply check against a threshold R θr.
Fig. 3 shows the DoG detector with and without edge suppression.
1.4 Detecting Extrema (10 pts)
To detect corner-like, scale-invariant interest points, the DoG detector chooses points
that are local extrema in both scale and space. Here, we will consider a point’s eight
neighbors in space and its two neighbors in scale (one in the scale above and one in the
scale below).
Q 1.4: Write the function :
locsDoG = getLocalExtrema(DoGPyramid, DoGLevels, PrincipalCurvature,
th contrast, th r)
This function takes as input DoGPyramid and DoGLevels from Section 1.2 and PrincipalCurvature
from Section 1.3. It also takes two threshold values, th contrast and th r. The threshold θc should remove any point that is a local extremum but does not have a Difference
of Gaussian (DoG) response magnitude above this threshold (i.e. |D(x, y, σ)| θc). The
5
threshold θr should remove any edge-like points that have too large a principal curvature
ratio specified by PrincipalCurvature.
The function should return locsDoG, an N × 3 matrix where the DoG pyramid
achieves a local extrema in both scale and space, and also satisfies the two thresholds.
The first and second column of locsDoG should be the (x, y) values of the local extremum
and the third column should contain the corresponding level of the DoG pyramid where
it was detected. (Try to eliminate loops in the function so that it runs efficiently.)
NOTE: In all implementations, we assume the x coordinate corresponds to
columns and y coordinate corresponds to rows. For example, the coordinate
(10, 20) corresponds to the (row 21, column 11) in the image.
1.5 Putting it together (5 pts)
Q 1.5: Write the following function to combine the above parts into a DoG detector:
locsDoG, GaussianPyramid = DoGdetector(im, sigma0, k, levels,
th contrast, th r)
The function should take in a gray scale image, im, scaled between 0 and 1, and
the parameters sigma0, k, levels, th contrast, and th r. It should use each of the
above functions and return the keypoints in locsDoG and the Gaussian pyramid in
GaussianPyramid. Figure 3 shows the keypoints detected for an example image. Note
that we are dealing with real images here, so your keypoint detector may find points
with high scores that you do not perceive to be corners.
Save the image with the detected keypoints as results/1 5.jpg and include it in
your report (similiar to the one shown in Fig. 3 (b) ). You can use any of the provided
images.
2 BRIEF Descriptor (35 pts)
Now that we have interest points that tell us where to find the most informative feature
points in the image, we can compute descriptors that can be used to match to other
views of the same point in different images. The BRIEF descriptor encodes information
from a 9 × 9 patch p centered around the interest point at the characteristic scale of the
interest point. See the lecture notes to refresh your memory.
HINT: All the functions to implement here are located in BRIEF.py.
2.1 Creating a Set of BRIEF Tests (5 pts)
The descriptor itself is a vector that is n-bits long, where each bit is the result of the
following simple test:
τ (P; x, y) := (
1, if P[x] < P[y].
0, otherwise.
(6)
6
Set n to 256 bits (R
256). Thus there are 256 τ . There is no need to encode the test
results as actual bits. It is fine to encode them as a 256 element vector.
There are many choices for the 256 test pairs {x, y} (remember x and y are 2D
vectors relating to discrete 2D coordinates within the 2D image patch matrix P) used
to compute τ (P; x, y) (each of the n bits). The authors describe and test some of them
in [2]. Read Section 3.2 of that paper and implement one of these solutions. You should
generate a static set of test pairs and save that data to a file. You will use these pairs
for all subsequent computations of the BRIEF descriptor.
Q 2.1: Write the function to create the x and y pairs that we will use for comparison
to compute τ :
compareX, compareY = makeTestPattern(patchWidth, nbits)
patchWidth is the width of the image patch (usually 9) and nbits is the number of
tests n in the BRIEF descriptor. compareX and compareY are linear indices into the
patchWidth × patchWidth image patch and are each nbits × 1 vectors. Run this routine for the given parameters patchWidth = 9 and n = 256 and save the results in
testPattern.npy. Include this file in your submission.
2.2 Compute the BRIEF Descriptor (10 pts)
Now we can compute the BRIEF descriptor for the detected keypoints.
Q 2.2: Write the function:
locs,desc = computeBrief(im, GaussianPyramid, locsDoG, k, levels,
compareX, compareY)
Where im is a grayscale image with values from 0 to 1, locsDoG are the keypoint
locations returned by the DoG detector from Section 1.5, levels are the Gaussian scale
levels that were given in Section 1, and compareX and compareY are the test patterns
computed in Section 2.1 and were saved into testPattern.npy.
The function returns locs, an m×3 vector, where the first two columns are the image
coordinates of keypoints and the third column is the pyramid level of the keypoints, and
desc is an m × n bits matrix of stacked BRIEF descriptors. m is the number of valid
descriptors in the image and will vary. You may have to be careful about the input DoG
detector locations since they may be at the edge of an image where we cannot extract
a full patch of width patchWidth. Thus, the number of output locs may be less than
the input locsDoG. Note: Its possible that you may not require all the arguments to
this function to compute the desired output. They have just been provided to permit
the use of any of some different approaches to solve this problem.
2.3 Putting it all Together (5 pts)
Q 2.3: Write a function :
locs, desc = briefLite(im)
7
which accepts a grayscale image im with values between 0 and 1 and returns locs, an
m × 3 vector, where the first two columns are the image coordinates of keypoints and
the third column is the pyramid level of the keypoints, and desc, an m × n bits matrix
of stacked BRIEF descriptors. m is the number of valid descriptors in the image and
will vary. n is the number of bits for the BRIEF descriptor.
This function should perform all the necessary steps to extract the descriptors from
the image, including
• Compute DoG pyramid.
• Get keypoint locations.
• Compute a set of valid BRIEF descriptors.
2.4 Check Point: Descriptor Matching (5 pts)
A descriptor’s strength is in its ability to match to other descriptors generated by the
same world point despite change of view, lighting, etc. The distance metric used to
compute the similarity between two descriptors is critical. For BRIEF, this distance
metric is the Hamming distance. The Hamming distance is simply the number of bits
in two descriptors that differ. (Note that the position of the bits matters.)
To perform the descriptor matching mentioned above, we have provided the function in
BRIEF.py):
matches = briefMatch(desc1, desc2, ratio)
Which accepts an m1 × n bits stack of BRIEF descriptors from a first image and a
m2×n bits stack of BRIEF descriptors from a second image and returns a p×2 matrix of
matches, where the first column are indices into desc1 and the second column are indices
into desc2. Note that m1, m2, and p may be different sizes and p ≤ min (m1, m2).
Q 2.4: Write a test script or utilize the code provided in the main function of BRIEF.py
to load two of the chickenbroth images and compute feature matches. Use the provided
plotMatches and briefMatch functions to visualize the result.
plotMatches(im1, im2, matches, locs1, locs2)
where im1 and im2 are two colored images stored as uint8, matches is the list of
matches returned by briefMatch and locs1 and locs2 are the locations of keypoints
from briefLite.
Save the resulting figure as results/2 4.jpg and submit it in your PDF report. Also,
present results with the two incline images and with the computer vision textbook cover
page (template is in file pf scan scaled.jpg) against the other pf * images. Briefly
discuss any cases that perform worse or better. See Figure 4 for an example result.
Suggestion for debugging: A good test of your code is to check that you can match
an image to itself.
8
Figure 4: Example of BRIEF matches for model chickenbroth.jpg and
chickenbroth 01.jpg.
2.5 BRIEF and rotations (10 pts)
You may have noticed worse performance under rotations. Let’s investigate this!
Q 2.5: Take the model chickenbroth.jpg test image and match it to itself while rotating the second image (hint: cv2.getRotationMatrix2D(...), cv2.warpAffine(...))
in increments of 10 degrees. Count the number of matches at each rotation and construct a bar graph showing rotation angle vs the number of correct matches. Include
this in your PDF and explain why you think the descriptor behaves this way. Write
your code in briefRotTest.py.
3 Planar Homographies: Theory (30 pts)
Suppose we have two cameras looking at a common plane in 3D space. Any 3D point
w on this plane generates a projected 2D point located at u˜ = [u1, v1, 1]T on the first
camera and x˜ = [x2, y2, 1]T on the second camera. Since w is confined to a plane, we
expect that there is a relationship between u˜ and x˜. In particular, there exists a common
3 × 3 matrix H, so that for any w, the following condition holds:
λx˜ = Hu˜ (7)
where λ is an arbitrary scalar weighting. We call this relationship a planar homography.
Recall that the ˜ operator implies a vector is employing homogenous coordinates such
9
that x˜ = [x, 1]T
. It turns out this homography relationship is also true for cameras that
are related by pure rotation without the planar constraint.
Q 3.1 (20 pts) We have a set of N 2D homogeneous coordinates {x˜1, . . . , x˜N } taken at
one camera view, and {u˜1, . . . , u˜N } taken at another. Suppose we know there exists an
unknown homography H between the two views such that,
λnx˜n = Hu˜n, for n = 1 : N (8)
where again λn is an arbitrary scalar weighting.
(a) Given the N correspondences across the two views and using Equation 8, derive a
set of 2N independent linear equations in the form:
Ah = 0 (9)
where h is a vector of the elements of H and A is a matrix composed of elements
derived from the point coordinates. Write out an expression for A.
Hint: Start by writing out Equation 8 in terms of the elements of H and the homogeneous coordinates for u˜n and x˜n.
(b) How many elements are there in h?
(c) How many point pairs (correspondences) are required to solve this system?
Hint: How many degrees of freedom are in H? How much information does each
point correspondence give?
(d) Show how to estimate the elements in h to find a solution to minimize this homogeneous linear least squares system. Step us through this procedure.
Hint: Use the Rayleigh quotient theorem (homogeneous least squares).
Q 3.2 Understanding homographies under rotation (5 points)
Suppose that a camera is rotating about its center C, keeping the intrinsic parameters
K constant. Let H be the homography that maps the view from one camera orientation
to the view at a second orientation. Let θ be the angle of rotation between the two.
Show that H2
is the homography corresponding to a rotation of 2θ. Please limit your
answer within a couple of lines. A lengthy proof indicates that you’re doing something
too complicated (or wrong).
Q 3.3 Limitations of the planar homography (2 points)
Why is the planar homography not completely sufficient to map any arbitrary scene
image to another viewpoint? State your answer concisely in one or two sentences.
Q 3.4 Behavior of lines under perspective projections (3 points)
We stated in class that perspective projection preserves lines (a line in 3D is projected
to a line in 2D). Verify algebraically that this is the case, i.e., verify that the projection
P in x = PX preserves lines.
10
4 Planar Homographies: Implementation (10 pts)
Note: Implement the method in planarH.py.
Now that we have derived how to find H mathematically in Q 3.1, we will implement
in this section.
Q 4.1 (10pts) Implement the function
H2to1 = computeH(X1,X2)
Inputs: X1 and X2 should be 2 × N matrices of corresponding (x, y)
T
coordinates between
two images.
Outputs: H2to1 should be a 3 × 3 matrix encoding the homography that best matches
the linear equation derived above for Equation 8 (in the least squares sense). Hint:
Remember that a homography is only determined up to scale. The numpy.linalg function
eigh() or svd() will be useful. Note that this function can be written without an explicit
for-loop over the data points.
5 RANSAC (15 pts)
Note: Implement the method in planarH.py.
The least squares method you implemented for computing homographies is not robust to
outliers. If all the correspondences are good matches, this is not a problem. But even a
single false correspondence can completely throw off the homography estimation. When
correspondences are determined automatically (using BRIEF feature matching for instance), some mismatches in a set of point correspondences are almost certain. RANSAC
(Random Sample Consensus can be used to fit models robustly in the presence of outliers.
Q 5.1 (15pts): Write a function that uses RANSAC to compute homographies automatically between two images:
bestH = ransacH(matches, locs1, locs2, nIter, tol)
The inputs and output of this function should be as follows:
• Inputs: locs1 and locs2 are matrices specifying point locations in each of the
images and matches is a matrix specifying matches between these two sets of
point locations. These matrices are formatted identically to the output of the
provided briefMatch function.
• Algorithm Input Parameters: nIter is the number of iterations to run RANSAC
for, tol is the tolerance value for considering a point to be an inlier. Define your
function so that these two parameters have reasonable default values.
• Outputs: bestH should be the homography model with the most inliers found
during RANSAC.
11
Figure 5: Harry potter cover warped to cv text book cover. This result is reproduced
with ransac max iter 20000 and tol 3.
6 Automated Homography Estimation and Warping (20
pts)
Q 6.1 Putting it together (10 points): Write a script HarryPotterize.py that
1. Reads pf desk.jpg, pf scan scaled.jpg, and hp cover.jpg.
2. Computes a homography automatically using briefLite, matchBrief, and ransacH.
3. Warps hp cover.jpg to the dimensions of the pf desk.jpg image using the OpenCV
function cv2.warpPerspective function.
4. At this point you should notice that although the image is being warped to the
correct location, it is not filling up the same space as the book. Why do you think
this is happening? How would you modify hp cover.jpg to fix this issue?
5. Implement the function in planarH.py:
composite img = compositeH(H, template, img)
to now compose this warped image with the desk image as in in Figure 5.
6. Save your image as results/6 1.jpg and include your resulting image in your
report. Please also print the final H matrix in your report.
Q 6.2 RANSAC Parameter Tuning (10 points) There are two tunable parameters in RANSAC which will we be exploring. Conduct a small ablation study by running
HarryPotterize.py with various max iters and inlier tol values. Include the result images
in your writeup, and explain the effect of these two parameters respectively. What is a
good selection for the number of random samples and why?
12
Figure 6.a:
incline L.jpg (img1)
Figure 6.b:
incline R.jpg (img2)
Figure 6.c: img2 warped to
img1’s frame
Figure 6: Example output for Q 7.1: Original images img1 and img2 (left and center)
and img2 warped to fit img1 (right). Notice that the warped image clips out of the
image. We will fix this in Q 7.2
7 Stitching it together: Panoramas (30 pts)
NOTE: All the functions to implement here are in panoramas.py.
We can also use homographies to create a panorama image from multiple views of the
same scene. This is possible for example when there is no camera translation between the
views (e.g., only rotation about the camera center). First, you will generate panoramas
using matched point correspondences between images using the BRIEF matching in
Section 2.4. We will assume that there is no error in your matched point
correspondences between images (Although there might be some errors, and
even small errors can have drastic impacts). In the next section you will extend
the technique to deal with the noisy keypoint matches.
You will need to use the perspective warping function from OpenCV: warp im =
cv2.warpPerspective(im, H, out size), which warps image im using the homography transform H. The pixels in warp_im are sampled at coordinates in the rectangle (0, 0)
to (out_size[0]-1, out_size[1]-1). The coordinates of the pixels in the source image
are taken to be (0, 0) to (im.shape[1]-1, im.shape[0]-1) and transformed according
to H. To understand this function, you may review Homework 0.
Q 7.1 (15pts) In this problem you will complete the partially implemented function:
panoImg = imageStitching(img1, img2, H2to1)
on two images from the Dusquesne incline. Please fill in the TODOs in the script. This
function accepts two images and the output from the homography estimation function.
This function:
(a) Warps img2 into img1’s reference frame using the aforementioned perspective warping function
(b) Blends img1 and warped img2 and outputs the panorama image.
For this problem, use the provided images incline L as img1 and incline R as img2.
The point correspondences pts are generated by your BRIEF descriptor matching.
13
Apply your ransacH() to these correspondences to compute H2to1, which is the
homography from incline R onto incline L. Then apply this homography to incline R
using warpH().
Note: Since the warped image will be translated to the right, you will need a larger
target image.
Visualize the warped image and save this figure as results/7 1.jpg using OpenCV’s
cv2.imwrite() function and save only H2to1 as results/q7 1.npy using Numpy np.save()
function.
Q 7.2 (15pts) Notice how the output from Q 7.1 is clipped at the edges? We will fix
this now. A partially implemented function is provided:
[panoImg] = imageStitching noClip(img1, img2, H2to1)
that takes in the same input types and produces the same outputs as in Q 7.1.
Please fill in the TODOs in the script.
To prevent clipping at the edges, we instead need to warp both image 1 and image 2
into a common third reference frame in which we can display both images without any
clipping. Specifically, we want to find a matrix M that only does scaling and translation
such that:
warp im1 = cv2.warpPerspective(im1, M, out size)
warp im2 = cv2.warpPerspective(im2, np.matmul(M,H2to1), out size)
This produces warped images in a common reference frame where all points in im1 and
im2 are visible. To do this, we will only take as input either the width or height of
out size and compute the other one based on the given images such that the warped
images are not squeezed or elongated in the panorama image. For now, assume we only
take as input the width of the image (i.e., out size[0]) and should therefore compute
the correct height(i.e., out size[1]).
Hint: The computation will be done in terms of H2to1 and the extreme points
(corners) of the two images. Make sure M includes only scale (find the aspect ratio of
the full-sized panorama image) and translation.
Test your method by passing incline L as img1 and incline R as img2.
Q 7.3 (0pts): This function has been provided to you. You now have all the tools you
need to automatically generate panoramas. This function accepts two images as input,
computes keypoints and descriptors for both the images, finds putative feature correspondences by matching keypoint descriptors, estimates a homography using RANSAC
and then warps one of the images with the homography so that they are aligned and
then overlays them.
im3 = generatePanorama(im1, im2)
Run your code on the image pairs {data/incline L.jpg, data/incline R.jpg} and
{data/hi L.jpg, data/hi R.jpg}. However during debugging, try on scaled down versions of the images to keep running time low. Save the resulting panorama on the full
sized images as results/q7 3.jpg. (see Figure 7 for example output). Include the
figure in your writeup.
14
Figure 7: Final panorama view. With homography estimated using RANSAC.
8 Extra Credit (40 pts)
The extra credit opportunities described below are optional and provide an avenue to
explore computer vision and improve the performance of the techniques developed above.
1. Host an online study group with at least 4 students. Up to 2 study groups are
counted for EC for each assignment. Each hosting is worth 5 points. (10 pts max)
2. You have attended Ce Liu’s VASC seminar or Tali Dekel’s AI seminar. Please
provide a half page report on what you learned from the seminar. (10 pts)
3. As we have seen, BRIEF is not rotationally invariant. Design a simple fix to solve
this problem using the tools you have developed so far. Explain in your PDF your
design decisions and how you selected any parameters that you use. Demonstrate
the effectiveness of your algorithm on image pairs related by large rotation. Submit
your code as ec 1.py. (10pts)
4. This implementation of BRIEF has some scale invariance, but there are limits.
What happens when you match a picture to the same picture at half the size?
Look to section 3 of [4] for a technique that will make your detector more robust
to changes in scale. Implement it and demonstrate it in action with several test
images. You may simply rescale some of the test images we have given you. Submit
your code as ec 2.py. (10pts)
For each question, plot a graph of the number of correct matches, changing rotation
from 0 to 180 degrees and scale from 1x to 20x (or some other set of appropriate scale
values) and analyze your results in your report.
15
References
[1] P. Burt and E. Adelson. The Laplacian Pyramid as a Compact Image Code. IEEE
Transactions on Communications, 31(4):532–540, April 1983.
[2] Michael Calonder, Vincent Lepetit, Christoph Strecha, and Pascal Fua. BRIEF :
Binary Robust Independent Elementary Features.
[3] Christopher G Harris, Mike Stephens, et al. A combined corner and edge detector.
In Alvey vision conference, volume 15, pages 10–5244. Citeseer, 1988.
[4] David G. Lowe. Distinctive Image Features from Scale-Invariant Keypoints. International Journal of Computer Vision, 60(2):91–110, November 2004.
16