This documents are a example to image transfromation functions using opencv and numpy on python.
Import dependences and input images. Input image is located on media/images/ directory.
import numpy as np
import matplotlib.pyplot as plt
import cv2original_image = cv2.imread("media/images/face.png", 0)
print(original_image.shape)(300, 300)
plt.figure(figsize=(4,4))
plt.imshow(original_image, cmap="gray")
plt.axis('off')
plt.show()
Illumination filter use gamma function to add brightness or darkness to image. If gamma value must be a real positive. If value is less to one, images is cleared, if value is higer to one, image is darkenet. If gamma is equal to 1, function return the same input image.
More info about gamma function here: https://docs.opencv.org/3.4/d3/dc1/tutorial_basic_linear_transform.html
Look this examplee:
from illumination import GammaFiltergf = GammaFilter()
gamma_values = [0.1, 0.25, 0.5, 0.75, 1, 2, 4, 8]config_dict = {"gamma":0}
plt.figure(figsize=(14,7))
for i, item in enumerate(gamma_values):
config_dict["gamma"] = item
gf.config(config_dict)
plt.subplot(2, 4, i + 1)
plt.title("gamma: {}".format(item))
plt.imshow(gf.filter(original_image), cmap="gray")
plt.axis('off')
plt.show()
Blur is another way to modify an picture. In this example we use three way to apply blur. AvgBlur set each pixel avergare value of your neighbors, MedianBlur apply an median instead of average and GaussianBlur use a gaussian kernel to blur image.
Neighbors used depend on kernel value, an kernel value high increase smoothing effect on image.
More info here: https://docs.opencv.org/master/d4/d13/tutorial_py_filtering.html
from blur import AvgBlur, GaussianBlur, MedianBlurfilters = [AvgBlur(), GaussianBlur(), MedianBlur()]
names = ["avg: ", "gauss: ", "median: "]
kernels = [3, 5, 7, 11]plt.figure(figsize=(3 * len(kernels), 3 * len(filters)))
plt.axis('off')
for i, f in enumerate(filters):
for j,k in enumerate(kernels):
c_config = {"kernel": k}
f.config(c_config)
output_image = f.filter(original_image)
plt.subplot(len(filters), len(kernels), j + 1 + i*len(kernels))
plt.title(names[i] + str(k))
plt.imshow(output_image, cmap="gray")
plt.axis('off')
plt.show()
Add noise is a effective way to modify a image. This examples apply Salt-Peper noise, Gaussian Noise, Sperckle noise.
Salt-Peper Noise set a random pixel to maximun and minimun values, in this case, 255 and 0.
from noise import SPNoise, GaussianNoise, PoissonNoisesp_values = [0.001, 0.005, 0.01, 0.05, 0.1, 0.5]
spnoiser = SPNoise()plt.figure(figsize=(3 * len(sp_values), 4))
for i, item in enumerate(sp_values):
spnoiser.config({"percent": item})
output_image = spnoiser.addNoise(original_image)
plt.subplot(1, len(sp_values), i + 1)
plt.title("sp percent: {}".format(item))
plt.imshow(output_image, cmap="gray")
plt.axis('off')
plt.show()
Gausian Noise add a random gaussian value to each pixel, this values are defined by a gaussian probabilty function with mean
gauss_noiser = GaussianNoise()
mean_values = [4, 8, 16, 32, 64]
sigma_values = [4, 8, 32, 64]
plt.figure(figsize=(4 * len(sigma_values), 4 * len(mean_values)))
for i, u in enumerate(mean_values):
for j,s in enumerate(sigma_values):
gauss_noiser.config({"mean": u, "sigma": s})
output_image = gauss_noiser.addNoise(original_image)
# print(output_image.shape)
plt.subplot(len(mean_values), len(sigma_values), j + 1 + i*len(sigma_values))
plt.title("mean: {}, sigma: {}".format(u,s))
plt.imshow(output_image, cmap="gray")
plt.axis('off')
plt.show()
Poisson Noise Poisson noise add to each pixel a radom value defined by poisson propabilty function.
pn = PoissonNoise()
output_image = pn.addNoise(original_image)plt.figure(figsize=(8,4))
plt.subplot(1,2,1)
plt.title("original")
plt.imshow(original_image, cmap="gray")
plt.axis('off')
plt.subplot(1,2,2)
plt.title("Poisson Noise")
plt.imshow(output_image, cmap="gray")
plt.axis('off')
plt.show()
Sperckle Noise multiply each pixel by a random value defined by gaussian probability funtion.
from noise import SepeckleNoise
sn = SepeckleNoise()
output_image = sn.addNoise(original_image)plt.figure(figsize=(8,4))
plt.subplot(1,2,1)
plt.title("original")
plt.imshow(original_image, cmap="gray")
plt.axis('off')
plt.subplot(1,2,2)
plt.title("SepeckleNoise")
plt.imshow(output_image, cmap="gray")
plt.axis('off')
plt.show()
A simple way to increase and reduce brightness and contrast of a images, is changing range of pixels function of an image. You can reduce distance betweeen maximun and minimun value, if you multiply each valur by a number between 0 and 1. If you need te opposite effect, number can be higher to 1. Also you can modidy offset of image.
from contrast import ABContrastabconstrast = ABContrast()
beta_values = [-128, -64, -32, 32, 64, 128]
alpha_values = [0.25, 0.5, 1, 2, 4]
plt.figure(figsize=(2.5 * len(beta_values), 2.5 * len(alpha_values)))
for i, u in enumerate(alpha_values):
for j,s in enumerate(beta_values):
abconstrast.config({"alpha": u, "beta": s})
output_image = abconstrast.filter(original_image)
plt.subplot(len(alpha_values), len(beta_values), j + 1 + i*len(beta_values))
plt.title("alpha: {}, beta: {}".format(u,s))
plt.imshow(output_image, cmap="gray")
plt.axis('off')
plt.show()
Another ways to change constract can be applying histogram equalization or using cubic and cubic root to reduce and increasse contrast, respectively.
from contrast import IncreaseContrast, ReduceContrast, EqulizerHistogramFilteric = IncreaseContrast()
rc = ReduceContrast()
eqc = EqulizerHistogramFilter()
ic_image = ic.filter(original_image)
rc_image = rc.filter(original_image)
equ_image = eqc.filter(original_image)
plt.figure(figsize=(14,4))
plt.subplot(1,4,1)
plt.title("original")
plt.imshow(original_image, cmap="gray")
plt.axis('off')
plt.subplot(1,4,2)
plt.title("increase contrast")
plt.imshow(ic_image, cmap="gray")
plt.axis('off')
plt.subplot(1,4,3)
plt.title("reduce contrast")
plt.imshow(rc_image, cmap="gray")
plt.axis('off')
plt.subplot(1,4,4)
plt.title("equalize histogram")
plt.imshow(equ_image, cmap="gray")
plt.axis('off')
plt.show()
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent