Week 4 - Optional programming exercises

In [7]:

    

%matplotlib inline

#import typical packages I'll be using
import cv2
import skimage
from skimage import data

import numpy as np
import matplotlib.pyplot as plt
import matplotlib.cm as cm
from pylab import rcParams
rcParams['figure.figsize'] = 16,8  #boiler plate to set the size of the figures
plt.gray()


    

    






<matplotlib.figure.Figure at 0x1a96c240>

In [2]:

    

img = data.camera()  #the cameraman image
width,height = img.shape

f, ax = plt.subplots(2,2,figsize=(24,24))

NOISE_PERCENT = [0, 0.05,.1,.2]
for c,noise in enumerate(NOISE_PERCENT):
    img_noisy = skimage.util.random_noise(img, mode = 's&p', amount = noise)*255

    ax[c/2,c%2].set_title('Cameraman Image & {:0.0%} Noise'.format(noise))
    ax[c/2,c%2].imshow(img_noisy);


    

In [3]:

    

f, ax = plt.subplots(2,2,figsize=(24,24))

FILTER_SIZE = [0,3,5,7]

noise = 0.05
img_noisy = skimage.util.random_noise(img, mode = 's&p', amount = noise)*255

for col,filter_size in enumerate(FILTER_SIZE): 
    img_recover = np.empty_like(img)

    f = filter_size/2

    if filter_size:
        for r in range(f,height-f):
            for c in range(f,width-f):
                #print r,c
                img_recover[r,c] = np.median(img_noisy[r-f:r+f,c-f:c+f])

    else:
        img_recover = img_noisy
        
    ax[col/2,col%2].set_title('{:0.0%} Noise - {}x{} Median Filter'.format(noise,filter_size,filter_size))
    ax[col/2,col%2].imshow(img_recover);


    

In [4]:

    

f, ax = plt.subplots(2,2,figsize=(24,24))

FILTER_SIZE = [0,3,5,7]

noise = 0.1
img_noisy = skimage.util.random_noise(img, mode = 's&p', amount = noise)*255

for col,filter_size in enumerate(FILTER_SIZE): 
    img_recover = np.empty_like(img)

    f = filter_size/2

    if filter_size:
        for r in range(f,height-f):
            for c in range(f,width-f):
                #print r,c
                img_recover[r,c] = np.median(img_noisy[r-f:r+f,c-f:c+f])

    else:
        img_recover = img_noisy
        
    ax[col/2,col%2].set_title('{:0.0%} Noise - {}x{} Median Filter'.format(noise,filter_size,filter_size))
    ax[col/2,col%2].imshow(img_recover);


    

In [5]:

    

f, ax = plt.subplots(2,2,figsize=(24,24))

FILTER_SIZE = [0,3,5,7]

noise = 0.2
img_noisy = skimage.util.random_noise(img, mode = 's&p', amount = noise)*255

for col,filter_size in enumerate(FILTER_SIZE): 
    img_recover = np.empty_like(img)

    f = filter_size/2

    if filter_size:
        for r in range(f,height-f):
            for c in range(f,width-f):
                #print r,c
                img_recover[r,c] = np.median(img_noisy[r-f:r+f,c-f:c+f])

    else:
        img_recover = img_noisy
        
    ax[col/2,col%2].set_title('{:0.0%} Noise - {}x{} Median Filter'.format(noise,filter_size,filter_size))
    ax[col/2,col%2].imshow(img_recover);


    

In [62]:

    

#Wiener example - http://scikit-image.org/docs/0.11.x/auto_examples/plot_restoration.html?highlight=wiener

from skimage import data, color, restoration
from scipy.signal import convolve2d as conv2

img = color.rgb2gray(data.astronaut())

psf = np.ones((5, 5)) / 25.
img_blur = conv2(img, psf, 'same')
img_blur += 0.1 * img_blur.std() * np.random.standard_normal(img_blur.shape)

deconvolved, _ = restoration.unsupervised_wiener(img_blur, psf)

f, ax = plt.subplots(1,2,figsize=(24,12))

ax[0].set_title('Blurry Image')
ax[0].imshow(img_blur)

ax[1].set_title('Wiener')
ax[1].imshow(deconvolved);


    

In [63]:

    

#Use the wiener method - choosing the parameter manually

f, ax = plt.subplots(1,2,figsize=(24,12))
deconvolved_img = restoration.wiener(img_blur, psf, .05)

ax[0].set_title('Blurry Image')
ax[0].imshow(img_blur)

ax[1].set_title('Wiener')
ax[1].imshow(deconvolved_img);


    

In [61]:

    

from skimage.restoration import nl_means_denoising

f, ax = plt.subplots(1,2,figsize=(24,12))
ax[0].set_title('Non-Local Means Denoising')
denoise = nl_means_denoising(img_blur, patch_size=5, patch_distance=11, h=25, fast_mode=True)
ax[0].imshow(denoise);

ax[1].set_title('Wiener')
ax[1].imshow(deconvolved_img);


    

In [76]:

    

#Using openCV to blur the image using a normalized block filter
f, ax = plt.subplots(2,3,figsize=(24,16))

blur = cv2.blur(img,(3,3))
denoise = nl_means_denoising(blur, patch_size=5, patch_distance=11, h=25, fast_mode=True)
ax[0,0].imshow(blur)
ax[0,0].set_title('3x3 kernel')
ax[1,0].set_title('Non-Local Means Denoising')
ax[1,0].imshow(denoise);

blur = cv2.blur(img,(10,10))
denoise = nl_means_denoising(blur, patch_size=5, patch_distance=11, h=25, fast_mode=True)
ax[0,1].imshow(blur)
ax[0,1].set_title('10x10 kernel')
ax[1,1].set_title('Non-Local Means Denoising')
ax[1,1].imshow(denoise);

blur = cv2.blur(img,(20,20))
denoise = nl_means_denoising(blur, patch_size=5, patch_distance=11, h=25, fast_mode=True)
ax[0,2].imshow(blur)
ax[0,2].set_title('20x20 kernel')
ax[1,2].set_title('Non-Local Means Denoising')
ax[1,2].imshow(denoise);


    

In [ ]:

    

TBD


    

In [ ]:

    

TBD


    

In [ ]:

    

TBD