def main(): # tryout image_list = listdir(IMAGE_PATH) im = import_pics(choice(image_list)) patch = get_patch(im,16) patch = normalize(patch) patches = get_patches(10, im, 16) patches = normalize_set(patches)
from os import listdir from random import choice import numpy as np import pylab as mpl from common import import_pics, show_image, IMAGE_PATH from 1.3 import get_patch, normalize_set # take 500 random images and slice out 16x16 patches image_list = listdir(IMAGE_PATH) patches = [] for i in range(500) im = import_pics(choice(image_list)) patches.append(get_patch(im,16)) patches = normalize_set(patches) # generate white noise images noises = [] for i in range(500) noises.append(np.random.normal(0, 15000, (16, 16))) normalize_set(noises) noises = normalize_set(noises) # generate smoothed white noise images fnoises = [] for i in range(500) noise = np.random.normal(0, 15000, (16, 16)) n = 9 fnoises.append(convolve2d(noise, np.ones((n, n)) / (n * n), mode="same", boundary="symm")) fnoises = normalize_set(fnoises)
import numpy as np
import pylab as mpl
from scipy.signal import convolve2d
from common import import_pics, show_image
# 1.2
# import an image
im = import_pics("imk01765.tiff")
# a)
# add white noise using different variances
for i in (100, 1000, 5000, 10000, 30000):
# generate white noise
whiten = np.random.normal(0, i, (1020, 1532))
# add the noise to the image
imnoise = im + whiten
# show image
show_image(imnoise, "Noise added (sigma=%d)" % i)
# b)
# generate white noise
whiten = np.random.normal(0, 15000, (1020, 1532))
kernels = (3, 5, 9, 15, 29)
for n in kernels:
# create smooth white noise by convolving white noise with a rectangular window
tmp = convolve2d(whiten, np.ones((n, n)) / (n * n), mode="same", boundary="symm")
show_image(im + tmp, "Noise kernel (%d,%d)" % (n, n))
# first manual approach (with 1-d vector)