Image Basics

Brief introduction of OpenCV
Pixels, colors, and image formats

OpenCV

OpenCV is an open source computer vision library written in C/C++ language
OpenCV contains more than 500 functions derived from various fields of computer vision
- Image segmentation
- Face recognition
- Action recognition
- Motion tracking
- Motion analysis
The image is treated as a matrix in the computer

import pandas as pd
import numpy as np

data_file = "./data/cat.csv"
cat = pd.read_csv(data_file)
cat

	R	G	B
0	213	218	222
1	213	218	222
2	213	218	222
3	213	218	222
4	213	218	222
...	...	...	...
243044	216	219	226
243045	216	219	226
243046	215	218	225
243047	215	218	225
243048	215	218	225

243049 rows × 3 columns

print(cat.shape)
print(type(cat))
# define matrix
width = height = 493
cat_rgb = []
for i in range(height):
    row = []
    for j in range(width):
        index = i * height + j
        rgb_element = [cat.at[index, 'R'], cat.at[index, 'G'], cat.at[index, 'B']]
        row.append(rgb_element)
    cat_rgb.append(row)
# data type transfermation
cat_rgb = np.array(cat_rgb)
print(cat_rgb.shape)
print(type(cat_rgb))

(243049, 3)
<class 'pandas.core.frame.DataFrame'>
(493, 493, 3)
<class 'numpy.ndarray'>

print(cat_rgb)
# 493 rows
# 493 rgb elements in each row

[[[213 218 222]
  [213 218 222]
  [213 218 222]
  ...
  [151 155 158]
  [150 154 157]
  [149 153 156]]

 [[213 218 222]
  [213 218 222]
  [213 218 222]
  ...
  [145 149 152]
  [144 148 151]
  [143 147 150]]

 [[213 218 222]
  [213 218 222]
  [213 218 222]
  ...
  [141 145 148]
  [140 144 147]
  [139 143 146]]

 ...

 [[ 19  18  14]
  [ 19  18  14]
  [ 19  18  14]
  ...
  [215 218 225]
  [215 218 225]
  [215 218 225]]

 [[ 19  18  14]
  [ 19  18  14]
  [ 18  17  13]
  ...
  [215 218 225]
  [215 218 225]
  [215 218 225]]

 [[ 19  18  14]
  [ 18  17  13]
  [ 18  17  13]
  ...
  [215 218 225]
  [215 218 225]
  [215 218 225]]]

from matplotlib import pyplot as plt
import matplotlib.colors as mat_color

no_norm = mat_color.Normalize(vmin=0, vmax=255, clip=False)
plt.imshow(cat_rgb, norm=no_norm)

<matplotlib.image.AxesImage at 0x1de53cd10a0>

png

Image formats

Just shown is the RGB image type, which is in line with human vision
In addition to the RGB series, common color spaces include HSV, HSL, XYZ, etc.
- Hue
- Saturation
- Value/Lightness
In image recognition, RGB is easily affected by light
- Manual compensation through programming
- Convert it into HSV mode
RGB -> HSV

\[s_{\mathrm{HSV}}=\frac{\max \{r, g, b\}-\min \{r, g, b\}}{\max \{r, g, b\}}\]

HSV -> RGB

\[\begin{aligned} c_{1} &=\left\lfloor h^{\prime}\right\rfloor \\ c_{2} &=h^{\prime}-c_{1} \\ w_{1} &=\left(1-s_{\mathrm{HSV}}\right) \cdot v \\ w_{2} &=\left(1-s_{\mathrm{HSV}} \cdot c_{2}\right) \cdot v \\ w_{3} &=\left(1-s_{\mathrm{HSV}} \cdot\left(1-c_{2}\right)\right) \cdot v \\ \left(\begin{array}{l} r \\ g \\ b \end{array}\right) &= \begin{cases}\left(v, w_{3}, w_{1}\right)^{\mathrm{T}} & \text { if } c_{1}=0 \\ \left(w_{2}, v, w_{1}\right)^{\mathrm{T}} & \text { if } c_{1}=1 \\ \left(w_{1}, v, w_{3}\right)^{\mathrm{T}} & \text { if } c_{1}=2 \\ \left(w_{1}, w_{2}, v\right)^{\mathrm{T}} & \text { if } c_{1}=3 \\ \left(w_{3}, w_{1}, v\right)^{\mathrm{T}} & \text { if } c_{1}=4 \\ \left(v, w_{1}, w_{2}\right)^{\mathrm{T}} & \text { if } c_{1}=5\end{cases} \end{aligned}\]

import cv2
import numpy as np
from matplotlib import pyplot as plt

path = "./images/cat.jpg"
# read original BGR image
img_bgr = cv2.imread(path)
print("image loaded")

image loaded

img_rgb = cv2.cvtColor(img_bgr, cv2.COLOR_BGR2RGB)
plt.imshow(img_rgb, norm=no_norm)

<matplotlib.image.AxesImage at 0x1de5406d9d0>

png

img_hsv = cv2.cvtColor(img_bgr, cv2.COLOR_RGB2HSV)
plt.imshow(img_hsv, norm=no_norm)

<matplotlib.image.AxesImage at 0x1de540d55b0>

png

Application of OpcnCV

Filtering, binarization, cutting, scale and rotation transformations, image gradients
Line and circle detection, feature point detection, edge detection, blob detection, feature point detection, pattern recognition
- QR code identification
- Face detection
- Gesture recognition
- Human gesture recognition