"""Image filters module.
All filters return a filtered copy of the image, the original image is
preserved.
"""
from PIL import Image, ImageFilter
import numpy as np
import pcf
[docs]def color_enhance(image):
"""Stretch all color channels to their full range."""
image_l = image.load()
min_r, min_g, min_b = 999, 999, 999
max_r, max_g, max_b = -1, -1, -1
for x in xrange(image.size[0]):
for y in xrange(image.size[1]):
min_r = min(min_r, image_l[x, y][0])
max_r = max(max_r, image_l[x, y][0])
min_g = min(min_g, image_l[x, y][1])
max_g = max(max_g, image_l[x, y][1])
min_b = min(min_b, image_l[x, y][2])
max_b = max(max_b, image_l[x, y][2])
new_image = Image.new('RGB', image.size)
new_image_l = new_image.load()
for x in xrange(image.size[0]):
for y in xrange(image.size[1]):
r, g, b = image_l[x, y]
r = (r - min_r) * 255 / (max_r - min_r)
g = (g - min_g) * 255 / (max_g - min_g)
b = (b - min_b) * 255 / (max_b - min_b)
new_image_l[x, y] = (r, g, b)
return new_image
[docs]def edge_detection(image):
"""Edge detection (on BW images)."""
new_image = image.filter(ImageFilter.GaussianBlur())
# GaussianBlur is undocumented class, it might not work in future versions
# of PIL
new_image = Image.frombytes('L', image.size,
pcf.edge(image.size, image.tobytes()))
return new_image
[docs]def peaks(image):
"""Peak filter (on BW images)."""
image_l = image.load()
new_image = Image.new('L', image.size)
new_image_l = new_image.load()
for x in range(2, image.size[0] - 2):
for y in range(2, image.size[1] - 2):
pix = (sum([sum([
- image_l[a, b]
for b in range(y - 2, y + 3)])
for a in range(x - 2, x + 3)])
+ (17 * image_l[x, y]))
if pix > 255:
pix = 255
if pix < 0:
pix = 0
new_image_l[x, y] = pix
return new_image
[docs]def high_pass(image, height):
"""High pass filter (on BW images)."""
image_l = image.load()
new_image = Image.new('L', image.size)
new_image_l = new_image.load()
for x in xrange(image.size[0]):
for y in xrange(image.size[1]):
if image_l[x, y] < height:
new_image_l[x, y] = 0
else:
new_image_l[x, y] = image_l[x, y]
return new_image
def components(image, diameter):
# TODO comment
# TODO refactor
image_l = image.load()
new_image_l = np.zeros(image.size, dtype=np.int)
components = [None]
comp_counter = 1
if diameter == 1:
for y in xrange(1, image.size[1] - 1):
for x in xrange(1, image.size[0] - 1):
if image_l[x, y]:
s = {0}
s.add(new_image_l[x - 1, y - 1])
s.add(new_image_l[x, y - 1])
s.add(new_image_l[x + 1, y - 1])
s.add(new_image_l[x - 1, y])
if len(s) == 1:
components.append(set())
new_image_l[x, y] = comp_counter
components[comp_counter].add((x, y))
comp_counter += 1
elif len(s) == 2:
s.remove(0)
c = s.pop()
new_image_l[x, y] = c
components[c].add((x, y))
else:
s.remove(0)
c1, c2 = s.pop(), s.pop()
components[c2].add((x, y))
for (x1, y1) in components[c2]:
new_image_l[x1, y1] = c1
components[c1] = components[c1] | components[c2]
components[c2] = None
elif diameter == 2:
for y in xrange(2, image.size[1] - 2):
for x in xrange(2, image.size[0] - 2):
if image_l[x, y]:
s = {0}
for (a, b) in [(a,b) for a in range(x - 2, x + 3)
for b in range(y - 2, y + 1)]:
if not (b == y and a >= x):
s.add(new_image_l[a, b])
if len(s) == 1:
components.append(set())
new_image_l[x, y] = comp_counter
components[comp_counter].add((x, y))
comp_counter += 1
elif len(s) == 2:
s.remove(0)
c = s.pop()
new_image_l[x, y] = c
try:
components[c].add((x, y))
except AttributeError:
print s, c
raise AttributeError
else:
s.remove(0)
c1 = s.pop()
components[c1].add((x, y))
new_image_l[x, y] = c1
for c2 in s:
for (x1, y1) in components[c2]:
new_image_l[x1, y1] = c1
components[c1] = components[c1] | components[c2]
components[c2] = None
else:
pass #TODO error
new_image = Image.new('L', image.size)
new_image_l = new_image.load()
for component in components:
if component:
x_c = 0
y_c = 0
c = 0
for (x, y) in component:
x_c += x
y_c += y
c += 1
new_image_l[int(round(float(x_c)/c)), int(round(float(y_c)/c))] = 255
return new_image