DEV Community: Rishi Saxena

Frequency Domain Filtering Using OpenCV

Rishi Saxena — Wed, 20 Oct 2021 09:15:10 +0000

What we need to get started with OpenCv...

We need to import few libraries given below and are available in Google Colab, independent installations may be required for other platforms.

1. Imports required



import cv2
from google.colab.patches import cv2_imshow
import numpy as np
import matplotlib.pyplot as plt

2. Next we import an image and use a simple Edge Preserving Filter.

Domain Filtering - Frequency Domain Filters are used for smoothing and sharpening of image by removal of high or low frequency components. Sometimes it is possible of removal of very high and very low frequency.



#read image
src = cv2.imread(r'/content/P.png', cv2.IMREAD_UNCHANGED)
DF = cv2.edgePreservingFilter(src, flags=1, sigma_s=60, sigma_r=0.4)
print("Domain Filter - Cartoonify")
cv2_imshow(numpy.hstack((src, DF)))

3. A- Smoothing of Image and other Domain Filters.

Frequency Domain Filters are used for smoothing and sharpening of images by removal of high or low-frequency components.
Gaussian blur (also known as Gaussian smoothing) is the result of blurring an image by a Gaussian function.



#apply guassian blur on src image
dst = cv2.GaussianBlur(src,(5,5),cv2.BORDER_DEFAULT)
print("Gaussian Smoothing")
cv2_imshow(numpy.hstack((src, dst)))

Mean Filter



kernel = np.ones((10,10),np.float32)/25   
dst2 = cv2.filter2D(src,-1,kernel)
print("Mean Filter")
cv2_imshow(numpy.hstack((src, dst2)))

Median Filter



#Median Filter
dst3 = cv2.medianBlur(src,5)
print("Median Filter")
cv2_imshow(numpy.hstack((src, dst3)))

4. B- Sharpening Filter -

Bilateral filter



#Bilateral filter 
print("Bilateral Filter")
dst4 = cv2.bilateralFilter(src, 60, 60, 60) 
cv2_imshow(numpy.hstack((src, dst4)))

5. C- Frequency Band Filter

Low Pass



#low pass filter
Lp = cv2.filter2D(src,-1, kernel)
Lp = src - Lp
print("Low Pass")
cv2_imshow(numpy.hstack((src, Lp)))

High Pass



Hp = src - dst
filtered = filtered + 127*numpy.ones(src.shape, numpy.uint8)
print("High Pass")
cv2_imshow(numpy.hstack((src, Hp)))

Frequency filters process an image in the frequency domain. The image is Fourier transformed, multiplied with the filter function and then re-transformed into the spatial domain. Attenuating high frequencies results in a smoother image in the spatial domain, attenuating low frequencies enhances the edges.

Any frequency filter can be implemented in the spatial domain and, if there exists a simple kernel for the desired filter effect.

If you missed the first part

LINK

OpenCV image processing using Python

Rishi Saxena ・ Sep 9 '21

#python #jupyter #opencv

Thanks for Reading...

OpenCV image processing using Python

Rishi Saxena — Thu, 09 Sep 2021 08:41:51 +0000

What we need to get started with OpenCv...

We need to import few libraries given below and are available in Google Colab, independent installations may be required for other platforms.

1. Imports required



from scipy.spatial import distance as dist
from imutils import perspective
from imutils import contours
import numpy as np
import argparse
import imutils
import cv2
import matplotlib.pyplot as plt
from google.colab.patches import cv2_imshow

2. Next we import an image and get its details



mg = cv2.imread(r'/content/parrot.jpg',cv2.IMREAD_UNCHANGED)
height = img.shape[0]  
width = img.shape[1]  
channels = img.shape[2]  
size1 = img.size
cv2_imshow(img)
print('Image Height       : ',height)  
print('Image Width        : ',width)  
print('Number of Channels : ',channels)  
print('Image Size  :', size1)

Remember we are using Colab and it uses its own snippets.

3. First lets try to get distance between two pixels



pixel = img[100,100]
pixel1 = img[200,200]
pixel_diff= pixel1-pixel  
print("The difference between the two pixels is :",pixel_diff)

4. Next lets try Point processing in the spatial domain on Image, Image Negatives and Power-Law (Gamma) Transformation.

Negative



print("Part A : Negative of the image")
plt.imshow(img)
plt.show()
# negative transformed image
color = ('b', 'g', 'r')    
plt.show()

Power-Law (Gamma) Transformation



print("Part B : Power Law ")
img = cv2.imread('/content/parrot.jpg', cv2.IMREAD_UNCHANGED)
gamma_two_point_two = np.array(255*(img/255)**2.2,dtype='uint8')

# Similarly, Apply Gamma=0.4 
gamma_point_four = np.array(255*(img/255)**0.4,dtype='uint8')

# Display the images in subplots
img3 = cv2.hconcat([gamma_two_point_two,gamma_point_four])
cv2_imshow(img3)

We used hconcat for displaying results together.

5. lets try some Point processing in the spatial domain.

Contrast stretching



print("Part C : Gray-level slicing, Contrast stretching")
img = cv2.imread('/content/parrot.jpg', cv2.IMREAD_UNCHANGED)
def pixelVal(pix, r1, s1, r2, s2):
    if (0 <= pix and pix <= r1):
        return (s1 / r1)*pix
    elif (r1 < pix and pix <= r2):
        return ((s2 - s1)/(r2 - r1)) * (pix - r1) + s1
    else:
        return ((255 - s2)/(255 - r2)) * (pix - r2) + s2
r1 = 70
s1 = 0
r2 = 140
s2 = 255
pixelVal_vec = np.vectorize(pixelVal)

# Apply contrast stretching.
contrast_stretched = pixelVal_vec(img, r1, s1, r2, s2)

print("Constrat Strethcing :")
# Save edited image.
cv2_imshow(contrast_stretched)

Gray-Level Slicing



class pointProcessing:
 def slicedGreyScale(self,image):
 # T1 and T2 Represent Lower and Upper Threshold Value
  T1 = 100
  T2 = 200
  h, w, c = img.shape
  img_thresh_back = np.zeros((h,w), dtype=np.uint8)
  for i in range(h):
    for j in range(w):
      if (T1 < image[i,j] and image[i,j] < T2):
        img_thresh_back[i,j]= 255
      else:
        img_thresh_back[i,j]= image[i,j]
  cv2_imshow(img_thresh_back)
pointObj= pointProcessing()
pointObj.slicedGreyScale(img)

6. Nearest neighbour Interpolation & Bilinear Interpolation.

Use of Average neighbour value and Bilinear



#Nearest neighbor Interpolation Using cv2.resize()Python
near_img = cv2.resize(img,None, fx = 2, fy = 2, interpolation = cv2.INTER_NEAREST)
cv2_imshow(near_img)

# Bilinear Interpolation 
bilinear_img = cv2.resize(img,None, fx = .5, fy = .5, interpolation = cv2.INTER_LINEAR)
cv2_imshow(bilinear_img)

7. Lets try other operations available in OpenCV

Arithmetic operations — Addition, Division
Logical Operations on Binary Image — XOR, NOT
Geometrical Operations — Rotation, Affine Translation
Statistical operations — Mean, Variance

Addition and Division -



print("A : Addition and Division :")
img1 = cv2.imread('/content/parrot.jpg')
img2 = cv2.imread('/content/bg.jpg')
dst = cv2.addWeighted(img1,0.3,img2,0.7,0)
#Div
div = cv2.divide(img1, img2)
AddDiv = cv2.hconcat([dst,div])
cv2_imshow(AddDiv)

XOR and NOT



print("B : Xor and Not Operations :")
#XOR function
bitwiseXor = cv2.bitwise_xor(img1, img2)
#NOT function
bitwiseNot = cv2.bitwise_not(img1)
#concat
img5 = cv2.hconcat([bitwiseXor,bitwiseNot])
cv2_imshow(img5)

Rotation and Affine Translation



print("C : Geometric Operations :")
print("Rotation and Affine Translation :")
#Rotation
image = cv2.rotate(img1, cv2.cv2.ROTATE_90_CLOCKWISE)
cv2_imshow(image)
#Affine Translation
srcTri = np.array( [[0, 0], [img1.shape[1] - 1, 0], [0, img1.shape[0] - 1]] ).astype(np.float32)
dstTri = np.array( [[0, img1.shape[1]*0.33], [img1.shape[1]*0.85, img1.shape[0]*0.25], [img1.shape[1]*0.15, img1.shape[0]*0.7]] ).astype(np.float32)
warp_mat = cv2.getAffineTransform(srcTri, dstTri)
warp_dst = cv2.warpAffine(img1, warp_mat, (img1.shape[1], img1.shape[0]))
# Rotating the image after Warp
center = (warp_dst.shape[1]//2, warp_dst.shape[0]//2)
angle = -50
scale = 0.6
rot_mat = cv2.getRotationMatrix2D( center, angle, scale )
warp_rotate_dst = cv2.warpAffine(warp_dst, rot_mat, (warp_dst.shape[1], warp_dst.shape[0]))
cv2_imshow(warp_dst)

Mean and Variance



print("D : Mean, Variance :")
#Mean of img1 and img2 
img7 = (img1+img2) * 0.5;
cv2_imshow(img7)
#Variance

Image interpolation : Down Sampling



print("E : Image interpolation : Down Sampling")
ds = cv2.pyrDown(img1)
cv2_imshow(ds)

As of now, We have covered the basics of OpenCV