In [1]:
import numpy as np # numerical python
from PIL import Image # Python Image Library
import matplotlib.pyplot as plt # displaying images
import matplotlib
import cv2
%matplotlib inline
1. Display Image and Image Depth¶
PIL, Matplotlib: BBG
OpenCV: BGR
In [2]:
# Reading image using PIL
img = Image.open('./data/test.jpg')
img
Out[2]:
In [3]:
# loading using matplotlib : (RBG Red Blue Green)
img_mat = matplotlib.image.imread("./data/test.jpg")
img_mat
Out[3]:
array([[[98, 52, 26],
[94, 48, 22],
[89, 43, 17],
...,
[11, 12, 0],
[13, 14, 0],
[14, 15, 1]],
[[95, 49, 23],
[94, 48, 22],
[91, 45, 19],
...,
[12, 13, 0],
[14, 15, 1],
[15, 16, 2]],
[[91, 45, 19],
[93, 47, 21],
[93, 47, 21],
...,
[13, 14, 0],
[14, 15, 1],
[15, 16, 2]],
...,
[[96, 66, 40],
[93, 63, 37],
[91, 61, 35],
...,
[73, 94, 0],
[76, 97, 2],
[76, 97, 2]],
[[94, 64, 38],
[91, 61, 35],
[91, 61, 35],
...,
[70, 91, 0],
[74, 95, 0],
[74, 95, 0]],
[[91, 61, 35],
[90, 60, 34],
[92, 62, 36],
...,
[70, 91, 0],
[76, 97, 2],
[75, 96, 1]]], dtype=uint8)
In [4]:
# read image using opencv : BGR (Blue Green and Red)
img_cv = cv2.imread('./data/test.jpg')
img_cv
Out[4]:
array([[[26, 52, 98],
[22, 48, 94],
[17, 43, 89],
...,
[ 0, 12, 11],
[ 0, 14, 13],
[ 1, 15, 14]],
[[23, 49, 95],
[22, 48, 94],
[19, 45, 91],
...,
[ 0, 13, 12],
[ 1, 15, 14],
[ 2, 16, 15]],
[[19, 45, 91],
[21, 47, 93],
[21, 47, 93],
...,
[ 0, 14, 13],
[ 1, 15, 14],
[ 2, 16, 15]],
...,
[[40, 66, 96],
[37, 63, 93],
[35, 61, 91],
...,
[ 0, 94, 73],
[ 2, 97, 76],
[ 2, 97, 76]],
[[38, 64, 94],
[35, 61, 91],
[35, 61, 91],
...,
[ 0, 91, 70],
[ 0, 95, 74],
[ 0, 95, 74]],
[[35, 61, 91],
[34, 60, 90],
[36, 62, 92],
...,
[ 0, 91, 70],
[ 2, 97, 76],
[ 1, 96, 75]]], dtype=uint8)
In [5]:
# display image
plt.imshow(img_mat) # rgb
Out[5]:
<matplotlib.image.AxesImage at 0x7f9430e712e0>
In [6]:
# display image
plt.imshow(img_cv)
Out[6]:
<matplotlib.image.AxesImage at 0x7f9450b80d00>
In [7]:
img_mat.shape
Out[7]:
(1200, 1600, 3)
In [8]:
# split image into three colors
# split array into three parts
r = img_mat[:,:,0] # red array
g = img_mat[:,:,1] # green array
b = img_mat[:,:,2] # blue array
r.shape,g.shape,b.shape
Out[8]:
((1200, 1600), (1200, 1600), (1200, 1600))
In [9]:
plt.figure(figsize=(10,6))
plt.subplot(1,3,1)
plt.title('Red region')
plt.imshow(r,cmap='gray')
plt.subplot(1,3,2)
plt.title('Green region')
plt.imshow(g,cmap='gray')
plt.subplot(1,3,3)
plt.title('Blue region')
plt.imshow(b,cmap='gray')
plt.show()
In [10]:
# extract information from the image
# gray scale : 2d (grayscale = 0.3 * red array + 0.59 * green array + 0.11 * blue array)
gray_MAT = cv2.cvtColor(img_mat,cv2.COLOR_RGB2GRAY)
# gray scale : 2d
gray_CV = cv2.cvtColor(img_cv,cv2.COLOR_BGR2GRAY)
gray_MAT.shape, gray_CV.shape
Out[10]:
((1200, 1600), (1200, 1600))
In [11]:
plt.imshow(gray_MAT,cmap='gray')
Out[11]:
<matplotlib.image.AxesImage at 0x7f944182d880>
In [12]:
plt.imshow(gray_CV,cmap='gray')
Out[12]:
<matplotlib.image.AxesImage at 0x7f9450f64520>
In [13]:
gray_CV
Out[13]:
array([[63, 59, 54, ..., 10, 12, 13],
[60, 59, 56, ..., 11, 13, 14],
[56, 58, 58, ..., 12, 13, 14],
...,
[72, 69, 67, ..., 77, 80, 80],
[70, 67, 67, ..., 74, 78, 78],
[67, 66, 68, ..., 74, 80, 79]], dtype=uint8)
2. Values and Pixels¶
- Pixels aka values range from $0$ to $(2^n - 1)$. For 8 bit image: $n = 8$. Range of pixels is $0$ (Black) to $255$ (White).
In [14]:
arr = np.arange(0,100,1)
arr1 = arr.reshape((10,10))
plt.imshow(arr1,cmap='gray')
Out[14]:
<matplotlib.image.AxesImage at 0x7f9471529520>
In [15]:
arr2 = np.random.randint(0,150,(10,10))
plt.imshow(arr2,cmap='gray')
Out[15]:
<matplotlib.image.AxesImage at 0x7f94315a6820>
In [16]:
img = cv2.imread('data/test.jpg') # bgr
gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
plt.imshow(gray,cmap='gray')
Out[16]:
<matplotlib.image.AxesImage at 0x7f9471486bb0>
In [17]:
slice = gray[0:10,0:10]
plt.imshow(slice,cmap='gray')
Out[17]:
<matplotlib.image.AxesImage at 0x7f94512c3ca0>
3. Resize¶
- Shrink -
cv2.INTER_AREA - Enlarge -
cv2.INTER_CUBIC(slow & good) orcv2.INTER_LINEAR(fast & ok)
In [18]:
# reading image
img = cv2.imread('./data/test.jpg') # bgr
# convert into grayscale
gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
print(gray.shape)
plt.imshow(gray,cmap='gray')
plt.show()
(1200, 1600)
In [19]:
# shrink (w=1600,h=1200)
# (w=500,h=400)
img_re = cv2.resize(gray,(500,400),cv2.INTER_AREA)
print(img_re.shape)
plt.imshow(img_re,cmap='gray')
plt.show()
(400, 500)
In [20]:
# englarge (w=2000, h = 1600)
img_en = cv2.resize(gray,(2000,1600),cv2.INTER_CUBIC)
print(img_en.shape)
plt.imshow(img_en,cmap='gray')
plt.show()
(1600, 2000)
4. Object Detection for Images (Face)¶
- Read Image
- Convert into Grayscale
- Apply Cascade Classifier
In [21]:
# read image
img = cv2.imread('./data/root_female.jpg')
# convert into gray scale
gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
# apply haar cascade classifier
haar = cv2.CascadeClassifier('./data/haarcascade_frontalface_default.xml')
# detect face
faces = haar.detectMultiScale(gray,1.3,5) # gray scale, scale size = 1.3, 5 neighbors
In [22]:
# [x, y, width, height] of the found face
print(faces)
[[127 55 205 205]]
In [23]:
cv2.rectangle(img,(127,55),(127+205,55+205),(0,255,0),3)
plt.imshow(img)
Out[23]:
<matplotlib.image.AxesImage at 0x7f94418a9b50>
In [24]:
# pop up the image
# cv2.imshow('object_detect',img)
# cv2.waitKey(0)
# cv2.destroyAllWindows()
In [25]:
# crop
face_crop = img[55:55+205,127:127+205]
plt.imshow(face_crop)
Out[25]:
<matplotlib.image.AxesImage at 0x7f94715b4610>
In [26]:
# extract the face from the image and save
cv2.imwrite('./data/root_female_extracted.png',face_crop)
Out[26]:
True
In [89]:
# read image
img = cv2.imread('./data/multi-people.jpg')
# convert into gray scale
gray = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
# apply haar cascade classifier
haar = cv2.CascadeClassifier('./data/haarcascade_frontalface_default.xml')
# detect face
faces = haar.detectMultiScale(gray,1.3,2) # gray scale, scale size = 1.3, 5 neighbors
In [90]:
print(gray.shape)
plt.imshow(gray,cmap='gray')
plt.show()
(200, 600)
In [91]:
print(faces)
for face in faces:
image = cv2.rectangle(img,(face[0],face[1]),(face[0]+face[2],face[1]+face[3]),(0,255,255),3)
plt.imshow(image)
[[ 46 19 38 38] [276 33 41 41] [129 42 37 37] [379 40 39 39] [537 40 38 38]]
5. Object Detection for Videos¶
In [93]:
# haar = cv2.CascadeClassifier('./data/haarcascade_frontalface_default.xml')
# def face_detect(img):
# gray = cv2.cvtColor(frame,cv2.COLOR_BGR2GRAY)
# faces = haar.detectMultiScale(gray,1.3,5)
# for x,y,w,h in faces:
# cv2.rectangle(img,(x,y),(x+w,y+h),(0,255,255),2)
# return img
# cap = cv2.VideoCapture('./data/video.mp4')
# while True:
# ret,frame = cap.read()
# if ret == False:
# break
# frame = face_detect(frame)
# cv2.imshow('object_detect',frame)
# #cv2.imshow('gray',gray)
# if cv2.waitKey(20) == 27:
# break
# cv2.destroyAllWindows()
# cap.release()
In [ ]:
In [ ]:
In [ ]: