intro to computer vision

1. Comput
er
Vision
by
Arefin Labib (1904111)

2. Concept of Image
21 45 12 56 77 122
.. .. .. .. .. .. ..
.. .. .. .. .. .. ..
.. .. .. .. .. .. ..
.. .. .. .. .. .. ..
.. .. .. .. .. .. ..
.. .. .. .. .. .. ..
visual representation of an object, scene, or information.

3. Reading Image
PIL OpenCV
Python Imaging Library.
Deprecated, PILLOW is
the improved version.

4. Types of Image
● Binary Image
Pixels are either black or white

5. Types of Image
● Binary Image
Pixels are either black or white
● Grayscale Image
Pixels range from 0 to 255

6. Grayscale vs BW image
255 0 255
0 255 0
0 0 255
255 18 220
25 201 33
7 45 255

7. Types of Image
● Binary Image
Pixels are either black or white
● Grayscale Image
Pixels range from 0 to 255
● RGB Image
Each pixel has 3 color channel

8. Pixel Value
● For BW Image
img_array[120, 344] = 0 or 255 and img_array is 2D array
● For Grayscale Image
img_array[120, 344] = ranged from 0 to 255 and img_array is 2D array
● For RGB Image
img_array[120, 344] = [x, y, z] ; x, y and z ranged from 0 to 255 and
img_array is 3D array

9. RED GREEN BLUE

10. PIL vs OpenCV
PIL: for basic image processing and manipulation
OpenCV: Designed for advanced computer vision and real-time image
processing
Video Processing is done with OpenCV

11. PIL vs OpenCV
PIL: Slower for large-scale processing
OpenCV: Optimized and faster, Can be used in C++ too
PIL: Uses RGB format
OpenCV: Uses BGR format by default

12. PIL vs OpenCV
PIL: Uses PIL Image objects
OpenCV: Uses NumPy arrays (numpy.ndarray)

13. Converting RGB to Grayscale
Simple Approach
Gray = (R+G+B)/3

14. Converting RGB to Grayscale
Maximum Method
Gray = max(R, G, B)

15. Converting RGB to Grayscale
Weighted Sum Method (Luminance Method)
Gray=0.2989×R + 0.5870×G + 0.1140×B

16. Brainstorm
Individual channels (R-channel, G-channel, B-channel) of
the images are -
A) Black and White
B) Grayscale
C) Will show specific color (Red only, Green only, Blue
only)

17. Brainstorm
Individual channels (R-channel, G-channel, B-channel) of
the images are -
A) Black and White
B) Grayscale
C) Will show specific color (Red only, Green only, Blue
only)

18. Basic Image Operations
● Resizing
resized_image = cv2.resize(img_array, (200, 100))
● Flipping
flip_vertical = cv2.flip(image, 0)
flip_horizontal = cv2.flip(image, 1)
flip_both = cv2.flip(image, -1)
● Cropping
cropped_image = image[50:200, 100:300]

19. Drawing Shapes in Image
● Circle
● Rectangle
● Line

20. Rectangle Drawing
cv2.rectangle(image, (50, 50), (250, 200), (0, 255, 0), 3)
cv2.rectangle(
image, # numpy.ndarray
(50, 50), # upper left pixel
(250, 200), # lower right pixel
(0, 255, 0), # color
3 thickness
)

21. Circle Drawing
cv2.circle(image, (350, 150), 50, (255, 0, 0), -1)
cv2.circle(
image,
(350, 150), # Center
50, # Radius
(255, 0, 0),
-1
)

22. Line Drawing
cv2.line(image, (100, 300), (400, 400), (0, 0, 255), 5)
cv2.line(
image,
(100, 300), # Starting point
(400, 400), # Ending point
(0, 0, 255),
5
)

23. Video Processing
import cv2
cap = cv2.VideoCapture("input_video.mp4")
while True:
ret, frame = cap.read()
cv2.imshow("Video", frame)
if cv2.waitKey(1) & 0xFF == ord('q'):
break
cap.release()
cv2.destroyAllWindows()

24. Video Processing
FPS: File/Frame per second
fps = cap.get(cv2.CAP_PROP_FPS)