Kuwahara filter in python
Project description
pykuwahara
Kuwahara filter in Python (numpy + OpenCV).
The Kuwahara filter is a non-linear smoothing filter used in image processing for adaptive noise reduction. It is able to apply smoothing on the image while preserving the edges. Source: Wikipedia
This implementation provide two variants of the filter:
- The classic one, using a uniform kernel to compute the window mean.
- A gaussian based filter, by computing the window gaussian mean. This is inspired by the ImageMagick approach.
Installation
pip install pykuwahara
Usage
Simple example
import cv2
from pykuwahara import kuwahara
image = cv2.imread('lena_std.jpg')
filt1 = kuwahara(image, method='mean', radius=3)
filt2 = kuwahara(image, method='gaussian', radius=3) # default sigma: computed by OpenCV
cv2.imwrite('lena-kfilt-mean.jpg', filt1)
cv2.imwrite('lena-kfilt-gaus.jpg', filt2)
Original image
Filtered with Kuwahara (mean)
Filtered with Kuwahara (gaussian)
Painting effect
Kuwahara filter can be used to apply a painting effet on pictures.
import cv2
from pykuwahara import kuwahara
image = cv2.imread('photo.jpg')
# Set radius according to the image dimensions and the desired effect
filt1 = kuwahara(image, method='mean', radius=4)
# NOTE: with sigma >= radius, this is equivalent to using 'mean' method
# NOTE: with sigma << radius, the radius has no effect
filt2 = kuwahara(image, method='gaussian', radius=4, sigma=1.5)
cv2.imwrite('photo-kfilt-mean.jpg', filt1)
cv2.imwrite('photo-kfilt-gaus.jpg', filt2)
Original image (source: wikipedia)
Filtered with Kuwahara (mean)
Filtered with Kuwahara (gaussian)
Advanced usage
Color image are supported by grayscaling the source image and using the gray channel to calculate the variance.
The user can provide another channel at his convenience, and alternatively give the right color conversion code (default is COLOR_BGR2GARY
).
import cv2
from pykuwahara import kuwahara
image = cv2.imread('selfie.jpg')
image = (image / 255).astype('float32') # pykuwahara supports float32 as well
lab_image = cv2.cvtColor(image, cv2.COLOR_BGR2Lab)
l, a, b = cv2.split(lab_image)
hsv_image = cv2.cvtColor(image, cv2.COLOR_BGR2HSV)
h, s, v = cv2.split(hsv_image)
filt1 = kuwahara(image, method='gaussian', radius=5, sigma=2., image_2d=l)
filt2 = kuwahara(image, method='gaussian', radius=5, sigma=2., image_2d=v)
cv2.imwrite('selfie-kfilt-gaus1.jpg', filt1 * 255)
cv2.imwrite('selfie-kfilt-gaus2.jpg', filt2 * 255)
Original image (source)
Filtered with Kuwahara on L (Lab)
Filtered with Kuwahara on V (HSV)
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Hashes for pykuwahara-0.3.2-py3-none-any.whl
Algorithm | Hash digest | |
---|---|---|
SHA256 | 79696b2c5b49a12cd603e946f0b6fb3572f96c333ce10fa4b559b083e872b41d |
|
MD5 | f2f4ac4aa4a9295c8814cbb53f6a67c9 |
|
BLAKE2b-256 | 530b11bbca85ffefe872284d98cbf6af1c010a11cf93e618496c97059bbfe914 |