%matplotlib inline

import matplotlib.pyplot as plt
import numpy as np

SIZE = (20,10)

from skimage.io import imread
from skimage import color, feature, filter

fig = plt.figure(figsize=SIZE)

a=fig.add_subplot(2,1,1)
plt.axis('off')

img = imread("/home/carpedm20/data/food100/1/1.jpg")
plt.imshow(img)

a=fig.add_subplot(2,1,2)
plt.axis('off')

img_gray = imread("/home/carpedm20/data/food100/1/1.jpg", flatten=True)

plt.imshow(img_gray, cmap='gray')
plt.show()

fig, ax1 = plt.subplots(1, 1, figsize=SIZE)

arr, hog = feature.hog(img_gray, 8, (16,16), (1,1), True)

ax1.axis('off')
ax1.imshow(img_gray, cmap=plt.cm.gray)
ax1.imshow(hog)

plt.show()

for y in img:
    for x in y:
        x=1
print img[0][0]

fig, ax1 = plt.subplots(1, 1, figsize=SIZE)

ax1.imshow(img)
ax1.imshow(hog)

plt.show()

x = np.array([[1, 2, 3], [4, 5, 6]], np.int32)
type(x)
x.shape
x.dtype
a = np.array([1,2,4,8,6,5,1,2,1]).reshape(3,3)
print a
a.max()
print a[0][1]
print a[0]
a[:]+=1
print a

img = imread("/home/carpedm20/data/food100/1/1.jpg")
type(img)
print img.shape

img_gray = imread("/home/carpedm20/data/food100/1/1.jpg", flatten=True)
print img_gray.shape

fig = plt.figure(figsize=SIZE)
plt.imshow(img)

print img[0][0]
new_img = img[::]-16
print new_img[0][0]

fig = plt.figure(figsize=SIZE)
plt.imshow(new_img)

img_gray = imread("/home/carpedm20/data/food100/1/1.jpg", flatten=True)
print img_gray.shape

size = (4,4)

plt.figure(1)
plt.imshow(img_gray[300:310,:10], cmap='gray', vmin=0, vmax=2, interpolation='nearest')
plt.colorbar()

print img_gray[300:305,:5]
new_img_gray = img_gray[:]+1
print new_img_gray[300:305,:5]

plt.figure(2)
plt.imshow(new_img_gray[300:310,:10], cmap='gray', vmin=0, vmax=2, interpolation='nearest')
plt.colorbar()

x = np.array([[1, 2, 3], [4, 5, 6]], np.int32)
fig = plt.figure(figsize=x.shape)
plt.imshow(x, cmap='gray')

print x.shape

plt.figure(1)
plt.imshow(x, cmap='gray', interpolation='nearest')
plt.grid(True)

print x
x[:]+=2
print x
fig = plt.figure(figsize=x.shape)

plt.figure(2)
plt.imshow(x, cmap='gray', interpolation='nearest')
plt.grid(True)

img_gray = imread("/home/carpedm20/data/food100/1/1.jpg", flatten=True)
print img_gray.shape

plt.figure(1)
plt.imshow(img_gray, cmap='gray', vmin=0, vmax=1, interpolation='nearest')
plt.colorbar()

print img_gray.max()

print img_gray[300:305,:5] 
new_img_gray = img_gray[:]*1/5 # linear transformation
print new_img_gray[300:305,:5]

plt.figure(2)
plt.imshow(new_img_gray, cmap='gray', vmin=0, vmax=1, interpolation='nearest')
plt.colorbar()

img_gray = imread("/home/carpedm20/data/food100/1/1.jpg", flatten=True)
print img_gray.shape

plt.figure(1)
plt.imshow(img_gray, cmap='gray', vmin=0, vmax=1, interpolation='nearest')
plt.colorbar()

print img_gray.max()

print img_gray[300:305,:5]
new_img_gray = np.fliplr(img_gray) # linear transformation
print new_img_gray[300:305,:5]

plt.figure(2)
plt.imshow(new_img_gray, cmap='gray', vmin=0, vmax=1, interpolation='nearest')
plt.colorbar()

from scipy import signal

img_gray = imread("/home/carpedm20/data/food100/1/1.jpg", flatten=True)

img_gray = img_gray[300:330,:30]

plt.figure(1)
plt.imshow(img_gray[2:-2,2:-2], cmap='gray', vmin=0, vmax=1, interpolation='nearest')
plt.colorbar()

f = np.ones((3,3)) # linear filter : blur (with mean filter)
f = f/f.size

print f

new_img_gray = signal.convolve(img_gray, f)

plt.figure(2)
plt.imshow(new_img_gray[2:-2,2:-2], cmap='gray', vmin=0, vmax=1, interpolation='nearest')
plt.colorbar()

from scipy import signal

img_gray = imread("/home/carpedm20/data/food100/1/1.jpg", flatten=True)

#img_gray = img_gray[300:330,:30]
img_gray = img_gray[200:300,:100]

plt.figure(1)
plt.imshow(img_gray[2:-2,2:-2], cmap='gray', vmin=0, vmax=1, interpolation='nearest')
plt.colorbar()

f = np.ones((3,3)) # linear filter : blur (with mean filter)
f = f/f.size

ff = np.array([[0,0,0],
               [0,2,0],
               [0,0,0]])
f = ff-f # linear filter : Sharpening filter

new_img_gray = signal.convolve(img_gray, f)

plt.figure(2)
plt.imshow(new_img_gray[2:-2,2:-2], cmap='gray', vmin=0, vmax=1, interpolation='nearest')
plt.colorbar()

from skimage import filter

img_gray = imread("/home/carpedm20/data/food100/1/1.jpg", flatten=True)

#img_gray = img_gray[300:330,:30]
img_gray = img_gray[200:300,:100]

plt.figure(1)
plt.imshow(img_gray[2:-2,2:-2], cmap='gray', vmin=0, vmax=1, interpolation='nearest')
plt.colorbar()

# Gaussian filter
# Removes "high-frequency" componets from the image (low-pass filter)

new_img_gray = filter.gaussian_filter(img_gray, sigma=1)
filter.gaussian_filter?

plt.figure(2)
plt.imshow(new_img_gray[2:-2,2:-2], cmap='gray', vmin=0, vmax=1, interpolation='nearest')
plt.colorbar()

from scipy import signal

img_gray = imread("/home/carpedm20/data/food100/1/1.jpg", flatten=True)

#img_gray = img_gray[300:330,:30]
img_gray = img_gray[200:300,:100]

plt.figure(1)
plt.imshow(img_gray[2:-2,2:-2], cmap='gray', vmin=0, vmax=1, interpolation='nearest')
plt.colorbar()

f = np.ones((3,3)) # linear filter : blur (with mean filter)
f = f/f.size

ff = np.array([[0,0,0],
               [0,2,0],
               [0,0,0]])

blur = signal.convolve(img_gray, f)

detail = img_gray[1:-1,1:-1] - blur[2:-2,2:-2]
print detail[0][:5]
print detail.max()

plt.figure(2)
plt.imshow(detail[2:-2,2:-2], cmap='gray', vmin=0, vmax=detail.max(), interpolation='nearest')
plt.colorbar()

new_img_gray = img_gray[1:-1,1:-1] + detail*2

plt.figure(3)
plt.imshow(new_img_gray[2:-2,2:-2], cmap='gray', vmin=0, vmax=1, interpolation='nearest')
plt.colorbar()


new_img_gray = img_gray[1:-1,1:-1] + detail*5

plt.figure(4)
plt.imshow(new_img_gray[2:-2,2:-2], cmap='gray', vmin=0, vmax=1, interpolation='nearest')
plt.colorbar()

from skimage.morphology import disk
#print disk(5)

gradient = filter.rank.gradient(img_gray, disk(3))

print gradient.max()

plt.figure(1)
plt.imshow(gradient, cmap='gray', interpolation='nearest')
plt.colorbar()

f = np.array([[-1,0,1],
              [-2,0,2],
              [-1,0,1]])

new_img_gray = signal.convolve(img_gray, f)

plt.figure(2)
plt.imshow(new_img_gray, cmap='gray', interpolation='nearest')
plt.colorbar()

f = np.array([[1,2,1],
              [0,0,0],
              [-1,-2,-1]])

new_img_gray = signal.convolve(img_gray, f)

plt.figure(3)
plt.imshow(new_img_gray, cmap='gray', interpolation='nearest')
plt.colorbar()

from skimage import data, color
img = color.rgb2gray(data.lena())

gradient = filter.rank.gradient(img, disk(3))

print gradient.max()

plt.figure(1)
plt.imshow(gradient, cmap='gray', interpolation='nearest')
plt.colorbar()

img = filter.gaussian_filter(img, sigma=1)

gradient = filter.rank.gradient(img, disk(3))

print gradient.max()

plt.figure(2)
plt.imshow(gradient, cmap='gray', interpolation='nearest')
plt.colorbar()

img = filter.gaussian_filter(img, sigma=3, mode='mirror')
#filter.gaussian_filter?
gradient = filter.rank.gradient(img, disk(3))

print gradient.max()

plt.figure(3)
plt.imshow(gradient, cmap='gray', interpolation='nearest')
plt.colorbar()


plt.figure(4)
plt.imshow(gradient, cmap='gray', interpolation='nearest')
plt.colorbar()