Description
Early Vision –One Image
For Homework 1, you will create a simple image-processing program, with
functions similar to those found in Adobe Photoshop or The Gimp. The functions you
implement for this assignment will take an input image, process the image, and
produce an output image.
Note: it would be possible to achieve some the functionality required in this
assignment by using built-in Matlab functions, especially from a couple of
specialized toolboxes. You are NOT allowed to use these image-specific built-in
functions. The convolution function (conv) is also not allowed. You will need to code
your own implementation of these functions. Other functions that are not allowed:
cart2pol, pol2cart. If in doubt, please ask if a certain built-in function is allowed.
Provided files:
A shell script vision_hwk1.m is provided to get you started. Also, a collection of
images is provided for testing. You can use your own images as well.
What You Have to Do
Implement a menu driven program, where each button should trigger a call to a
function. If you run the provided shell program, you will notice the following three
buttons:
1. Load Image – loads one image file. In order to select one image, the file
needs to be in the same folder as the main script. The loaded image will
become the current image and can be passed as an input to other
functions.
2. Display Image – displays the current image.
3. Exit Program – closes the menu and terminates the script.
You have to add and test additional functionality for the program. For every
button/functionality you add, you can use one of the images to test it.
Implement a solution for each of the following tasks and add a menu button for each
one. The solution/code for each task should be written as a separate function. Note:
in Matlab, every function is written in a separate file.
Task 1 (0.25 points) Brighten: individually modifies the RGB channels of an image,
brightening or darkening it. Use Loops!
function [ outImg ] = makeBright_L( inImg, brightness )
This function brightens each pixel and writes out the new image to outImg. The
value of the input parameter brightness is the amount by which the image
should be brightened, so it can be any real number in the range [-255 255]. Positive
values will brighten the image, while negative values will darken the image. You
must use loops to modify each individual pixel value.
Task 2 (0.5 points) Brighten2: individually modifies the RGB channels of an image,
brightening or darkening it. NO Loops!
function [ outImg ] = makeBright_NL( inImg, brightness )
This function will achieve the same thing as makeBright_L, but without the use
of loops.
• Add menu buttons for the makeBright_L and makeBright_NL
functionality. Choosing one of these two menu items should result in:
a. Asking the user to input a value between -255 and 255 for the brightness
parameter, followed by a call to the makeBright_L / NL function, passing the current
image and the value of brightness as inputs.
b. Displaying the original image and the resulting image, side by side (use subplots)
c. Save the resulting image.
Task 3 (0.25 points) Invert: inverts the colors of an image. Use Loops!
function [ outImg ] = invert_L( inImg)
Color inversion, also known as the negative effect, is one of the easiest effects to
achieve in image processing. Subtract each RGB color value from the maximum
possible value (255) and the result will be an inverted image. You must use loops to
modify each individual pixel values.
Choosing the Invert_L menu button should result in:
a. Calling the invert_L function with the current image as an input
b. Displaying the original image and the resulting image, side by side (use subplots)
c. Saving the resulting image.
Task 4 (0.5 points) Invert: inverts the colors of an image. NO Loops!
function [ outImg ] = invert_NL( inImg)
This function will achieve the same thing as invert_L, without the use of loops.
Choosing the Invert_NL menu button should result in:
a. Calling the invert_NL function with the current image as an input
b. Displaying the original image and the resulting image, side by side (use subplots)
c. Saving the resulting image.
Task 5 (0.5 points) AddRandomNoise: adds random noise to an image. No Loops!
function [ outImg ] = addRandomNoise_L( inImg)
The input image inImg has pixel values between 0 and 255 in all three channels R,G
and B. This function adds random noise to each pixel and writes out the new image
to outImg. The random noise added should be different for every pixel and it
should be in the range [-255, 255].
Choosing the AddRandomNoise_L menu button should result in:
a. Calling the addRandomNoise_L function, with the current image as input.
b. Displaying the original image and the resulting image, side by side (use subplots)
c. Saving the resulting image.
Task 6 (0.5 points) Luminance: change a color image into a luminance (gray scale)
image. No Loops!
function [ outImg ] = luminance_L( inImg)
This method transforms a color image into a gray image and writes out the new
image to outImg. Note: here, the variable outImg will be a 2-dimensional matrix.
You must use loops to modify each individual pixel values for outImg.
In order to convert a particular color into gray scale we need to figure out what the
intensity or brightness of that color is. A quick way is to calculate the mean value of
the red, green and blue channel components:
Although it can produce reasonable results, this method is not perfect. This is
because the human eye does not perceive reds, greens and blues at the same
intensity level. The color green, for example, at maximum intensity looks brighter
than blue at maximum intensity.
For this assignment we will use the following
weighting system:
I = 0.299R + 0.587G + 0.114B
Given this image:
Converting to gray scale using the mean method:
… and using the weighted method:
Choosing the Luminance_L menu button should result in:
a. Calling the luminance_L function, with the current image as input.
b. Displaying the original image and the resulting image, side by side (use subplots)
c. Saving the resulting image.
Task 7 (2 points) A red filter will make red things lighter and complementary
colors darker. That’s why a red filter, on black and white film, makes clouds stand
out — it doesn’t do a thing to the white clouds but it makes the blue sky darker, so
the clouds stand out against the darker sky tones.
ModifytheGreenandBluechannelsinthesamemanner.
The outImg matrix has the same size as the input image matrix inImg. The
middle 1/3 of the outImg image matrix is identical to the original, the left 1/3 is the
gray scale version of the original image, and the right 1/3 has a red filter applied.
Just as the description above implies, a red filter will make the red components
brighter and the blue and green components darker. The variable redVal can have
any value between 0 and 1. Here is how the redVal value is used in applying the
red filter. First, remember the weighting we used for each of the three channels in
the luminance function:
I = 0.299R + 0.587G + 0.114B
When applying the red filter, the result is a gray scale image, where the weight value
for the Red component will be redVal, and the Green and Blue components will
equally “share” the remaining weight. The sum of the three weights for Red, Green
and Blue should be equal to 1.
Choosing the Red Filter menu button should result in:
a. Asking the user to input a value for redVal between 0 and 1, followed by calling
the redFilter function, passing the current image and the redVal value as
inputs.
b. Displaying the original image and the resulting image, side by side (use subplots)
c. Saving the resulting image.
Task 8 (2.5 points) Binary image: similar to the blue-screening technique on TV
and in movies
function [ outImg] = binaryMask(inImg)
Create one binary image outImg (also known as the mask), in which the pixels
which belong to the “object of interest” will be white (1), and the background will be
black (0). The binaryMask function takes a grayscale image as an input. You can
call one of the luminance functions if the current image is a color one (sully.bmp is a
good choice), or you can use the wrench1.jpg image file.
Try to come up with an
algorithm that finds the threshold value that separates the background from the
“object of interest”. (if they are part of the wrench, in this example).
Choosing the Binary Image menu button should result in:
a. Calling the binaryMask function, with the current image as input.
b. Displaying the original image and the resulting image, side by side (use subplots)
c. Saving the resulting image.
Task 9 (5 points) Mean Filter: also known as smoothing, averaging or box filter
function [ outImg ] = meanFilter( inImg, kernel_size )
Mean filtering is a method of smoothing images, i.e. reducing the amount of intensity
variation between one pixel and the next. It is often used to reduce noise in images.
The idea of mean filtering is simply to replace each pixel value in an image with the
mean (average) value of its neighbors, including itself.
This has the effect of
eliminating pixel values that are unrepresentative of their surroundings.
The input argument kernel_size will determine the size of the smoothing
kernel. For example, if kernel_size is 3, the smoothing kernel is of 3 x 3 size like
in the picture below.
Note: Pay close attention to the pixels on the edge (first row, last row, first column,
last column). How many neighboring pixels do they have? Use selection statements
to address these special cases or pad the image with extra rows and columns
(details in lecture, remember the NaN value!).
Choosing the Mean Filter menu button should result in:
a. Ask the user to input the size of the smoothing kernel (a positive integer)
b. Call the meanFilter function, with the current image as input, plus the value
entered by the user for the size of the smoothing kernel.
c. Display the original image and the image returned by the function, side by side
(use subplots)
d. Save the resulting image. Use a naming convention of your choice.
Task 10 (3 points) Frosty Filter
function [ outImg ] = frosty( inImg, n, m )
This method applies a filter to the image similar to the frosted glass effect.
Simply replace each pixel value in the image with a random value from one of its
neighbors, including self, in an n by m window.
Choosing the Frosty Filter menu button should result in:
a. Ask the user to input a positive value for n, within the bounds of the image size
b. Ask the user to input a positive value for m, within the bounds of the image size
b. Call the frosty function, with the current image as input, plus the values entered
by the user for n and m.
c. Display the original image and the image returned by the function, side by side
(use subplots)
d. Save the resulting image. Use a naming convention of your choice.
Note: You might want to implement two helper functions useful for tasks 11-13:
function [ value ] = sampleNearest( x, y)
This method returns the value of the image, sampled at position (x,y) using nearestpoint sampling. https://en.wikipedia.org/wiki/Nearest-neighbor_interpolation
function [ value ] = sampleBilinear( x, y)
This method returns the value of the image, sampled at position (x,y) using bilinearweighted sampling. https://en.wikipedia.org/wiki/Bilinear_interpolation
Task 11 (5 points) Scale Nearest
function [ outImg ] = scaleNearest( inImg, factor )
This method scales an image using nearest point sampling to obtain pixel values
and returns the new image.
The value of the input parameter factor should be positive and it represents the
factor by which the height and width of the image are to be scaled. For example, if
the value of factor is 2, the width of the new image should be twice the width of
the original image; same with the height. If the value of factor is less than 1, for
example 0.3, then the new width will be 0.3 * the_value_of_the_original_width. If
effect, a value of factor less than 1 will result in a smaller image than the original.
Choosing the Scale Nearest menu button should result in:
a. Ask the user to input a positive value for factor
b. Call the scaleNearest function, with the current image as input, plus the value
entered by the user for factor.
c. Display the original image and the image returned by the function, side by side
(use subplots)
d. Save the resulting image. Use a naming convention of your choice.
Task 12 (7 points) Scale Bilinear
function [ outImg ] = scaleBilinear( inImg, factor )
This method scales an image using bilinear-interpolation to obtain pixel values and
returns the new image. The value of the input parameter factor should be positive
and it represents the factor by which the height and width of the image are to be
scaled. (see more examples at Task 3)
Choosing the Scale Bilinear menu button should result in:
a. Ask the user to input a positive value for factor
b. Call the scaleBilinear function, with the current image as input, plus the
value entered by the user for factor.
c. Display the original image and the image returned by the function, side by side
(use subplots)
d. Save the resulting image. Use a naming convention of your choice.
Task 13 (8 points) Swirl Filter
function [ outImg ] = swirlFilter( inImg, factor, ox, oy)
This method applies a swirl filter to the current image and returns the new image.
The swirl filter is a warp or a distortion. In the distorted image, the pixel from
location (r,c) in the original image is rotated an angle theta with respect to the origin
coordinates ox & oy, and it will end up at a new coordinate pair (x,y). Pixels closer
to the origin will rotate less, while pixles further from the origin will rotate more.
The input parameter factor determines the magnitude of the rotation, and the
direction. A positive value of factor will create a clockwise swirl, while a negative
value of factor should create a counter-clockwise swirl. You task is to inverse map
each (x,y) coordinate pair in the final image to the original (r,c) value. You can use a
non-linear mapping of your choice (bilinear, nearest-neighbor).
Choosing the Fun Filter menu button should result in:
a. Ask the user to enter values for factor, ox, oy. Make sure each of them are
within the allowed range of values.
b. Call the funFilter function, with the current image and the additional 3
parameters as input.
b. Displaying the original image and the image returned by the function, side by side
(use subplots).
c. Save the resulting image. Use a naming convention of your choice.
Task 14 (10 points) Famous Me
Create a composite image of yourself and a famous tourist attraction. Example:
“crop and paste” yourself on the moon. For this task you will need to use two
images: the original image (of yourself, ideally in front of a non-busy background,
either very light in color of very dark) and the target image (of the moon, for
example).
For the “crop” part, you will need to create a binary mask image using the
image of yourself. Then, use the binary image and select the location in pixel
coordinates where you want to “paste” yourself. If you want to resize the picture of
yourself (or the mask) with respect to the target, you can use one of the functions
you already developed for scaling.
Choosing the Famous Me menu button should result in:
a. Call the famousMe function. Note: You get to decide how you want to implement
this task, how many inputs you want your function to have and if it will you use
other already developed functions (binary mask, for example).
b. Display the 3 images side by side: original image, image of you and the resulting
image (use subplots)
c. Save the resulting image. Use a naming convention of your choice.
Submitting the assignment:
Make sure each script or function file is well commented and it includes a block
comment with your name, course number, assignment number and instructor name.
Save one resulting image for each of the buttons/functionality implemented. Zip all
the .m files and the image files together and submit the resulting .zip file through
Canvas under assignment H1, by Tuesday, January 26st, by 9:55pm
