Answered step by step
Verified Expert Solution
Link Copied!

Question

1 Approved Answer

1 . Create a new matlab script harrisCorner . m 2 . Read in the Penguins.jpg image from the Windows sample pictures folder using the

1. Create a new matlab script harrisCorner.m
2. Read in the Penguins.jpg image from the Windows sample pictures folder using the imread command (store it in myImage)
3. Convert the image to grayscale using the rgb2gray command.
4. Cast your variable to a double and reassign it to itself.
myImage = double(myImage)
5. Create the x and y derivative filters
dxFilter =[-10-1; -101; -101];
dyFilter = dxFilter';
6. Create an X derivative image and a Y derivative image by convolving the appropriate filters with the grayscale penguin image. Use the following names for your new images:
myImageDerivativeX
myImageDerivativeY
7. Now that we have the derivatives, we can calculate A, B, C from the book.
A = myImageDerivativeX .^2
B = myImageDerivativeY .^2
C = myImageDerivativeX .* myImageDerivativeY
8. After you calculate the above you want to apply Gaussian smoothing (using a Gaussian Filter)
using the conv2 command. Store each of the smoothed derivatives in: smoothedA, smoothedB,
smoothedC. You can use the following command to create your Gaussian filter:
gaussianFilter = fspecial('gaussian');
9. In order to compute the Corner Response Function, we use the following formula (define alpha=0.04):
cornerResponseFunction = A .* B C2- alpha *(A + B)2
**Note: Please be sure to use the smoothed versions of A, B & C for the above formula
10. Next we need to create our isLocalMax function. The Corner Response function usually will have several corners all that appear in the same area. This function will help us to determine which is the best corner for our program to use. Use the following code for isLocalMax:
function [myBool]= isLocalMax(harrisMatrix, u, v)
height = size(harrisMatrix,1);
width = size(harrisMatrix,2);
if(u <=1|| u >= height || v<=1|| v>=width)
myBool = false;
else
pix = reshape(harrisMatrix, height*width,1); %return the image as
a 1 dimensional array (like the book does)
i0=(v-1)*height+u;
i1= v*height+u;
i2=(v+1)*height+u;
cp = pix(i1);
myBool =(cp > pix(i0-1) && cp > pix(i0) && cp > pix(i0+1) && cp >
pix(i1-1) && cp > pix(i1+1) && cp > pix(i2-1) && cp > pix(i2) && cp >
pix(i2+1));
end
end
11. We use this function to help us determine the best corners to use. We do this by stepping
through our cornerResponseFunction matrix and check to see if its value is greater than a
certain threshold and it is a local maximum (this is checked by calling the isLocalMax function
you just created)
12. Implement the pseudo code listed below in order to perform the above task:
Initialize totalCorners to zero
For each u:
For each v:
If (cornerResponseFunction (u, v)> threshold && islocalMax(cornerResponseFunction, u, v)
Increment totalCorners by 1.
Add the corner to our list (each corner contains u, v, q)
cornerList(totalCorners).x = u;
cornerList(totalCorners).y = v;
cornerList(totalCorners).q = q;
Output the total number of corners that passed the threshold and isLocalMax
*NOTE: Use an initial threshold of 200000(threshold =200000)
13. Sort the corners by their respective q value in descending order
14. Create a new variable goodCorners and initialize it to an empty array:
goodCorners =[];
15. Our corner response function may return a lot of corners that can be quite close to each other.
What wed like to do next is define a minimum distance that we want these features (corners) to
be from each other. Well then want to step through each of our corners and remove the ones
that are close too close to each other. We do this by starting with the strongest corner that we
found (the first one in the list because its sorted) and remove other corners that are within the
minimum distance to this point.
16. First thing we will want to do is initialize our minimum distance:
minDistance =10;
17. Copy and paste the following code into your program and complete the code
while(size(cornerList,2)>0)
c1= cornerList(1);
cornerList = cornerList(2:end); %remove the first one
goodCorners =[goodCorners c1];
cornersToRemove =[];
for i=2:size(cornerList,2)
if(<<>>>)
cornersToRemove =[cornersToRemove i];
end
end
cornerList(cornersToRemove)=[]; %remove the corners that we found
that are too close
end
18. Step through the goodCorners list and mark each of the corners in the image with a green cross that is 3 pixels in width and height.
19. Display the image to the user

Step by Step Solution

There are 3 Steps involved in it

Step: 1

blur-text-image

Get Instant Access to Expert-Tailored Solutions

See step-by-step solutions with expert insights and AI powered tools for academic success

Step: 2

blur-text-image

Step: 3

blur-text-image

Ace Your Homework with AI

Get the answers you need in no time with our AI-driven, step-by-step assistance

Get Started

Recommended Textbook for

OpenStack Trove

Authors: Amrith Kumar, Douglas Shelley

1st Edition

1484212215, 9781484212219

More Books

Students also viewed these Databases questions

Question

How to sketch the process layout for a event planning buffet

Answered: 1 week ago

Question

4. How would you deal with the store manager?

Answered: 1 week ago