DV1614: Basic Edge Detection Using Python

DV1614: Basic edge detection using Python
August 31, 2020
1 Introduction to edge detection

An image consists of pixels, where each pixel is typically represented by three
values corresponding to three channels: red, green, and blue. Detecting edges
in an image is a frequently employed effect (see Figure 1) in many applications
of image processing, machine vision, computer vision, etc.
(a) Input image (b) After edge detection
Figure 1: Performing edge detection on an image
There are multiple ways to detect edges in an image. For edge detection,
the input image is typically converted to grayscale. An 8-bit grayscale image
represents a pixel using one value (between 0 and 255). The output image can
then be obtained by applying a simple technique on all the pixels of the input
image. Figure 2 illustrates this technique to detect edges in an image. The mask
is used to show large differences in pixel values. If a pixel’s neighboring pixels
differ strongly in intensity, its updated value will appear white, otherwise black.
1.1 Masks that can be used for detecting edges

Mathematical formulae can be used to derive a mask of arbitrary size using
which edge detection can be performed on the input image. However, to keep
1
things simple, we will use two predefined 3x3 masks, either of which can be used
to finish this task.
   
−1 0 1 −1 −1 −1
Mask 1 = −2 0 2 Mask 2 = −1 8 −1
   
−1 0 1 −1 −1 −1
Input image pixel values
1 2 0 7 0 3 5 0
8 7 5 4 1 0 1 0 Example 3x3 mask that can be
0 3 5 3 0 1 3 0 used for edge detection
1 2 1 7 1 0 7 5 -1 -1 -1
0 0 0 0 0 1 0 3 -1 8 -1
0 0 1 7 2 0 1 2 -1 -1 -1
Per-element mult. Output image

new pixel values
-7 -5 -4
-3 40 -3 Σ 8
-2 -1 -7 Add all values
Figure 2: Illustration of edge detection using an example mask. Source:

http://cg.ivd.kit.edu
2
Figure 3: Performing edge detection using the image in Figure 1 and Mask 2.
Source: https://setosa.io/ev/image-kernels/
2 Tasks to complete
This section describes the tasks constituting the edge detection program. In
order to complete this assignment, you need to complete the tasks 3–5 as per
the following instructions.
In the edgedetect assgn.py file:
1. Read the input image file and convert it to 8-bit grayscale format. In this
format, each pixel is represented using an integer in the range [0, 256).
2. Add dummy pixels with a value of 0 along the image boundary, so that
the 3x3 mask does not fall outside the actual image.
The above tasks are already implemented in the function read colorimg()
in the helper functions.py file given to you. The function returns
a list of lists representing the pixel values. Let us name this 2D list
pixel values.
3. Using list comprehension, create a list of lists to store the updated pixel
information. Let us name this new pixel values. The length of the list
is numb rows and the length of each sublist is numb colns. Initialize all
the values to 0.
4. Represent the values comprising the 3x3 image mask using a tuple of
tuples. Let us name this 2D tuple mask. You can use the values in Mask
1 or Mask 2.
5. Use mask to calculate a new value for every pixel in the image, excluding
the dummy pixels. Since pixel values is a 2D data structure, use a
3
nested for loop to iterate through every pixel. For each pixel, the following
are the tasks that you have to implement:
(a) Call a function get slice 2d list() (to be implemented by you),
which takes as input the pixel values, the position of the pixel being
calculated, and returns the 3x3 patch of surrounding pixels as a list
of lists. For example, when updating the pixel shown in Figure 2,
calling this function from within the nested for loop should return
neighbor pixels = [[7, 5, 4], [3, 5, 3], [2, 1, 7]].
Use list slicing on pixel values to extract the required neighboring
pixels. Use list comprehension to create the 2D list neighbor pixels.
(b) Call a function flatten() (to be implemented by you), which takes
as input a 2D list or a 2D tuple, and returns a 1D list. For example,
when this function is given [[7, 5, 4], [3, 5, 3], [2, 1, 7]] as input, it
should return [7, 5, 4, 3, 5, 3, 2, 1, 7]. When it is given ((-1, -1, -1),
(-1, 8, -1), (-1, -1, -1)), it returns [-1, -1, -1, -1, 8, -1, -1, -1, -1]. Using
this function, flatten neighbor pixels and mask.
Use list comprehension in the flatten() function to create a 1D list
from the 2D list.
(c) Apply the 1D list containing the mask values on the 1D list containing
the neighborhood pixels using the map() function and a lambda.
This lambda should multiply the corresponding elements of the two
flattened lists. Your implementation should look as follows:
map (lambda ..., list1, list2)
Note: This may not be the most efficient way to perform element-
wise multiplication of the 3x3 patch of pixels and the mask. However,
it is important to learn how to use the map() function.
(d) Use the output of map() function and sum all the multiplication
results to obtain the new value for the pixel. See the Σ operation in
Figure 2.
(e) Set the new value by appropriately indexing new pixel values. Re-
member that the 2D list new pixel values does not has dummy
pixels.
6. Verify your result against the correct result (computed using the scipy
package). If your computations are correct, “True result” will be printed
on the console. Otherwise, it will output “False result”.
7. View the original image and the detected edges.
The last two steps have been implemented in the functions verify result()
and view images() in the helper functions.py file.
4
3 Final note
Note that to be able to use the helper functions.py file, you need to install
the following Python packages: matplotlib, scikit-image, scipy, numpy.
After successful installation, when you run the code provided to you without
making any changes, it should look like this:
Figure 4: Screenshot of the output when the provided code is executed
You do not need the temp variable that currently exists in the code. You
can delete it when writing your code. After successfully finishing your tasks,
you will be able to see the edges of the input image in the window. Also, the
console will print a “True result” message.
If you have any questions, please send a mail to [email protected].
All the best!

DV1614: Basic Edge Detection Using Python

Uploaded by

Copyright:

Available Formats

DV1614: Basic Edge Detection Using Python

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

DV1614: Basic Edge Detection Using Python

Uploaded by

Copyright:

Available Formats

DV1614: Basic edge detection using Python

August 31, 2020

1 Introduction to edge detection

(a) Input image (b) After edge detection

Figure 1: Performing edge detection on an image

1.1 Masks that can be used for detecting edges

Input image pixel values

Per-element mult. Output image

Figure 2: Illustration of edge detection using an example mask. Source:

Figure 4: Screenshot of the output when the provided code is executed

You might also like