def gauss_filter():
header, y_cb_cr = bmp.readBmp(gauss_noise_filename)
h, y_cb_cr_o = bmp.readBmp(original_file_y_cb_cr)
sigmaRange = np.arange(0.0001, 1.1, 0.1)
psnr = defaultdict(list)
rRange = range(1, 10)
legends = []
out_psnr = np.zeros((len(list(rRange)) + 1, (len(list(sigmaRange)) + 1)))
out_psnr[1:, 0] = list(rRange)
out_psnr[0, 1:] = list(sigmaRange)
print('orig ', bmp.psnr(y_cb_cr_o[..., 0], y_cb_cr[..., 0]))
for r in rRange:
psnr = []
for i in sigmaRange:
res = BmpFilters.BmpFilters.gauss_filter(y_cb_cr, r, i)
ps = bmp.psnr(y_cb_cr_o[..., 0], res[..., 0])
res[..., 1] = res[..., 0]
res[..., 2] = res[..., 0]
bmp.writeBmp(
gauss_filter_filename +
"/gauss r = {0:.2f} d = {1:.2f}.bmp".format(r, i), header, res)
psnr.append(ps)
p = np.array(psnr)
out_psnr[r, 1:] = p
l, = plt.plot(list(sigmaRange), psnr, label=f'R = {r}')
legends.append(l)
plt.legend(handles=legends)
plt.savefig(gauss_filter_filename + f"/psnr total.png")
np.savetxt(gauss_filter_filename + '/psnr.csv',
out_psnr,
delimiter=',',
fmt='%.2f')
def median_for_impulse():
header, y_orig = bmp.readBmp(original_file_y_cb_cr)
folder = 'impulse2/'
folder_res = 'impulse2/res/'
pa = [0.02, 0.05, 0.2, 0.3]
pb = [0.03, 0.05, 0.05, 0.2]
for a, b in zip(pa, pb):
percent = (a + b) * 100
h, y = bmp.readBmp(folder + f'impulse % = {percent}.bmp')
psnr = []
x = []
labels = []
print(f'original {percent}', bmp.psnr(y_orig, y))
x.append(0)
psnr.append(bmp.psnr(y_orig, y))
for i in range(1, 10):
res = BmpFilters.BmpFilters.median_filter(y, i)
p = bmp.psnr(y_orig[..., 0], res[..., 0])
x.append(i)
psnr.append(p)
res[..., 1] = res[..., 0]
res[..., 2] = res[..., 0]
bmp.writeBmp(folder_res + f' % {percent} r = {i}.bmp', header, res)
labels.append(plt.plot(x, psnr, label=f'r = {i}'))
plt.savefig(folder + 'psnr.png', handlers=labels)
def gamma():
folder = 'C:\\Users\\HawkA\\PycharmProjects\\bmp-stats\\gamma\\'
lena_low = 'lenaLow.bmp'
lena_high = 'lenaHigh.bmp'
lena = 'lena.bmp'
mid_folder = 'mid\\'
low_folder = 'low\\'
high_folder = 'high\\'
h, y_high = bmp.readBmp(folder + lena_high)
h, y_low = bmp.readBmp(folder + lena_low)
h, y_orig = bmp.readBmp(folder + lena)
y, x = make_hist(y_high[..., 0])
plt.bar(x, y)
plt.savefig(folder + 'original high.png')
plt.clf()
y, x = make_hist(y_low[..., 0])
plt.bar(x, y)
plt.savefig(folder + 'original low.png')
plt.clf()
y, x = make_hist(y_orig[..., 0])
plt.bar(x, y)
plt.savefig(folder + 'original.png')
plt.clf()
# d = np.array(list(range(0,255)))
# plt.plot(d, BmpFilters.BmpFilters.gamma(d,1,25))
# plt.show()
y_rng = np.arange(1, 25, 0.5)
for i in y_rng:
res_high = BmpFilters.BmpFilters.gamma(y_high, 1, i)
res_low = BmpFilters.BmpFilters.gamma(y_low, 1, i)
orig = BmpFilters.BmpFilters.gamma(y_orig, 1, i)
bmp.writeBmp(folder + high_folder + f'lena y = {i:.2f}.bmp', h,
res_high)
bmp.writeBmp(folder + low_folder + f'lena y = {i:.2f}.bmp', h, res_low)
bmp.writeBmp(folder + mid_folder + f'lena y = {i:.2f}.bmp', h, orig)
y, x = make_hist(res_high[..., 0])
plt.bar(x, y)
plt.savefig(folder + high_folder + f"high {i:.2f}.png")
plt.clf()
y, x = make_hist(res_low[..., 0])
plt.bar(x, y)
plt.savefig(folder + low_folder + f"low {i:.2f}.png")
plt.clf()
y, x = make_hist(orig[..., 0])
plt.bar(x, y)
plt.savefig(folder + mid_folder + f"orig {i:.2f}.png")
plt.clf()
def hist():
folder = 'C:\\Users\\HawkA\\PycharmProjects\\bmp-stats\\hist\\'
lena_low = 'lenaLow.bmp'
lena_high = 'lenaHigh.bmp'
lena = 'lena.bmp'
h, y_high = bmp.readBmp(folder + lena_high)
h, y_low = bmp.readBmp(folder + lena_low)
h, y_orig = bmp.readBmp(folder + lena)
y_high_hist, x = make_hist(y_high[..., 0])
plt.bar(x, y_high_hist)
plt.savefig(folder + 'original high.png')
plt.clf()
y_low_hist, x = make_hist(y_low[..., 0])
plt.bar(x, y_low_hist)
plt.savefig(folder + 'original low.png')
plt.clf()
y_mid_hist, x = make_hist(y_orig[..., 0])
plt.bar(x, y_mid_hist)
plt.savefig(folder + 'original.png')
plt.clf()
look_up_table_mid = BmpFilters.BmpFilters.create_look_up(y_mid_hist)
look_up_table_low = BmpFilters.BmpFilters.create_look_up(y_low_hist)
look_up_table_high = BmpFilters.BmpFilters.create_look_up(y_high_hist)
res_orig_hist = BmpFilters.BmpFilters.hist(look_up_table_mid, y_orig)
res_high_hist = BmpFilters.BmpFilters.hist(look_up_table_high, y_high)
res_low_hist = BmpFilters.BmpFilters.hist(look_up_table_low, y_low)
y_mid_hist, x = make_hist(res_orig_hist[..., 0])
plt.bar(x, y_mid_hist)
plt.savefig(folder + 'mid2 hist.png')
plt.clf()
y_mid_hist, x = make_hist(res_high_hist[..., 0])
plt.bar(x, y_mid_hist)
plt.savefig(folder + 'high2 hist.png')
plt.clf()
y_mid_hist, x = make_hist(res_low_hist[..., 0])
plt.bar(x, y_mid_hist)
plt.savefig(folder + 'low2 hist.png')
plt.clf()
bmp.writeBmp(folder + "org2.bmp", h, res_orig_hist)
bmp.writeBmp(folder + "low2.bmp", h, res_low_hist)
bmp.writeBmp(folder + "high2.bmp", h, res_high_hist)
def median():
header, y_cb_cr = bmp.readBmp(gauss_noise_filename)
header_o, y_cb_cr_o = bmp.readBmp(original_file_y_cb_cr)
psnr = []
print('orig', bmp.psnr(y_cb_cr_o[..., 0], y_cb_cr[..., 0]))
x = range(1, 10)
for r in x:
res = BmpFilters.BmpFilters.median_filter(y_cb_cr, r * 2 + 1)
res[..., 1] = res[..., 0]
res[..., 2] = res[..., 0]
psnr.append(bmp.psnr(y_cb_cr_o[..., 0], res[..., 0]))
bmp.writeBmp(median_folder + f"/median r = {r}.bmp", header, res)
np.savetxt(median_folder + "/psnr.csv",
np.array(psnr).T,
delimiter=',',
fmt='%.2f')
plt.plot(list(x), psnr)
plt.savefig(median_folder + "/psnr.png")
def sobel():
folder = "sobel/"
# file = "test.bmp"
header, y_cb_cr = bmp.readBmp(original_file_y_cb_cr)
x = np.arange(0, 255, 5)
# for i in x:
y_cb_cr_max, angle_map = BmpFilters.BmpFilters.sobel(y_cb_cr, 127)
y_cb_cr_max[..., 1] = y_cb_cr_max[..., 0]
y_cb_cr_max[..., 2] = y_cb_cr_max[..., 0]
# bmp.writeBmp(folder + "sobel thr = {:d}.bmp".format(127), header, y_cb_cr_max)
bmp.writeBmp(folder + "angle.bmp", header, angle_map)
def max_freq():
header, y_cb_cr = bmp.readBmp(original_file_y_cb_cr)
y_cb_cr_max = BmpFilters.BmpFilters.max_freq(y_cb_cr)
y_cb_cr_max[..., 1] = y_cb_cr_max[..., 0]
y_cb_cr_max[..., 2] = y_cb_cr_max[..., 0]
y, x = make_hist(y_cb_cr[..., 0])
y2, x2 = make_hist(y_cb_cr_max[..., 0])
plt.bar(x, y)
plt.show()
plt.bar(x2, y2)
plt.show()
bmp.writeBmp(laplas_folder + "/max.bmp", header, y_cb_cr_max)
def lapslas_with_aplha():
header, y_cb_cr = bmp.readBmp(original_file_y_cb_cr)
header, max_y_cb_cr = bmp.readBmp(laplas_folder + "/max.bmp")
y, x = make_hist(max_y_cb_cr[..., 0])
plt.bar(x, y)
plt.savefig(laplas_folder + "/max freq .png")
plt.clf()
return
psnr = []
bright = []
y, x = make_hist(y_cb_cr[..., 0])
bright.append(np.average(y_cb_cr))
# plt.bar(x, y)
# plt.savefig(laplas_folder + "/original_hist.png")
# plt.clf()
x = np.arange(0, 1.6, 0.1)
for i in x:
y_cb_cr_max = BmpFilters.BmpFilters.laplas_operator(y_cb_cr, i)
bright.append(np.average(y_cb_cr_max))
y_cb_cr_max[..., 1] = y_cb_cr_max[..., 0]
y_cb_cr_max[..., 2] = y_cb_cr_max[..., 0]
y, x = make_hist(y_cb_cr_max[..., 0])
# plt.bar(x, y)
# plt.savefig(laplas_folder + "/hist alpha {:.1f}.png".format(i))
# plt.clf()
# psnr.append(bmp.psnr(y_cb_cr_max[..., 0], max_y_cb_cr[..., 0]))
# bmp.writeBmp(laplas_folder + "/alpha {:.1f}.bmp".format(i), header, y_cb_cr_max)
x = np.arange(0, 1.6, 0.1)
x = list(x)
x.insert(0, -1)
x = np.array(x, dtype=np.float)
y = np.array(bright)
print(len(x))
print(len(y))
np.savetxt(laplas_folder + "/bright average.csv", (x.T, y.T),
delimiter=',',
fmt='%.2f')
def laplas():
header, rgb = bmp.readBmp(orig_filename)
y_cb_cr = bmp.convertYCbCr(rgb)
y_cb_cr_res = BmpFilters.BmpFilters.laplas_operator(y_cb_cr, 0)
y_cb_cr_res += 128
np.putmask(y_cb_cr_res, y_cb_cr_res < 0, 0)
np.putmask(y_cb_cr_res, y_cb_cr_res > 255, 255)
y_cb_cr_res[..., 1] = y_cb_cr_res[..., 0]
y_cb_cr_res[..., 2] = y_cb_cr_res[..., 0]
y_cb_cr[..., 1] = y_cb_cr[..., 0]
y_cb_cr[..., 2] = y_cb_cr[..., 0]
bmp.writeBmp(laplas_folder + "/laplas.bmp", header, y_cb_cr_res)
bmp.writeBmp(laplas_folder + "/orig.bmp", header, y_cb_cr)
def impulse_noise():
# hist = defaultdict(list)
# x = defaultdict(list)
# hist[0.1].append(1)
# hist[0.1].append(2)
# x[0.1].append(3)
# x[0.1].append(6)
# x[0.1].append(6)
# for (k, v), (k2, v2) in zip(hist.items(), x.items()):
# print(k, v, k2, v2)
# print('\n')
#
# return
small = 0.00001
file_name = orig_filename
header, rgb = bmp.readBmp(file_name)
y_cb_cr = bmp.convertYCbCr(rgb)
hist = defaultdict(list)
x = defaultdict(list)
rng = np.arange(0, 1.1, 0.1)
for i in rng:
for j in rng:
if i + j >= 1:
continue
if i == j == 0:
i = small
j = small
res = BmpFilters.BmpFilters.add_impulse_noise(y_cb_cr, i, j)
p = bmp.psnr(y_cb_cr[..., 0], res[..., 0])
if i == j == small:
i = 0
j = 0
x[i].append(j)
hist[i].append(p)
res[..., 1] = res[..., 0]
res[..., 2] = res[..., 0]
bmp.writeBmp(
'impulse/impulse pa = {0:.1f} pb = {1:.1f}.png'.format(
float(i), float(j)), header, res)
l = []
for (k, v), (k1, v1) in zip(x.items(), hist.items()):
leg, = plt.plot(list(v),
list(v1),
label='pa = {0:.1}'.format(float(k)))
l.append(leg)
plt.legend(handles=l)
plt.savefig('impulse/psnr.png')
def impulse_with_persent():
folder = 'impulse2/'
header, y_cb_cr = bmp.readBmp(original_file_y_cb_cr)
pa = [0.02, 0.05, 0.2, 0.3]
pb = [0.03, 0.05, 0.05, 0.2]
psnr = []
x = []
for a, b in zip(pa, pb):
res = BmpFilters.BmpFilters.add_impulse_noise(y_cb_cr, a, b)
psnr.append(bmp.psnr(y_cb_cr[..., 0], res[..., 0]))
x.append((a + b) * 100)
res[..., 1] = res[..., 0]
res[..., 2] = res[..., 0]
bmp.writeBmp(folder + f'impulse % = {(a + b) * 100}.bmp', header, res)
plt.plot(x, psnr)
plt.savefig(folder + 'psnr.png')
def gauss_noise():
header, rgb = bmp.readBmp(orig_filename)
y_cb_cr = bmp.convertYCbCr(rgb)
psnr = []
rng = range(0, 250, 10)
for i in rng:
if i == 0:
i = 0.0000001
res = BmpFilters.BmpFilters.add_gauss_noise(y_cb_cr, d=i, m=0)
res[..., 1] = res[..., 0]
res[..., 2] = res[..., 0]
p = bmp.psnr(y_cb_cr[..., 0], res[..., 0])
psnr.append(p)
if i == 0.0000001:
i = 0
bmp.writeBmp(f"gauss/gauss d ={i}.bmp", header, res)
plt.plot(list(rng), psnr)
plt.savefig(gauss_noise_psnr_filename)
def avg():
file_name = gauss_noise_filename
header, rgb = bmp.readBmp(file_name)
psnr = []
rng = range(1, 25)
y_cb_cr = bmp.convertYCbCr(rgb)
orinal_y_cb_cr = bmp.convertYCbCr(orig_pic)
for i in rng:
res = BmpFilters.BmpFilters.avg(y_cb_cr, i)
psnr.append(bmp.psnr(orinal_y_cb_cr[..., 0], res[..., 0]))
res[..., 1] = res[..., 0]
res[..., 2] = res[..., 0]
# bmp.writeBmp(moving_average_folder + f"/avg {i}.bmp", header, res)
x = np.array(list(rng), dtype='uint')
y = np.array(psnr)
np.savetxt(moving_average_folder + "/psnr.csv", (x, y),
delimiter=',',
fmt='%.2f')
plt.plot(list(rng), psnr)
plt.savefig(moving_average_folder + "/psnr.png")
def white_and_black_generation():
folder = 'C:\\Users\\HawkA\\PycharmProjects\\bmp-stats\\bright\\'
high_bright_file = 'lenaHigh.bmp'
low_bright_file = 'lenaLow.bmp'
high_coeff = 2.0
low_coeff = 0.2
header, high_bright = bmp.readBmp(folder + high_bright_file)
y_high, x_high = make_hist(high_bright[..., 0])
plt.bar(x_high, y_high)
plt.savefig(folder + 'original_high_hist.png')
plt.clf()
header, low_bright = bmp.readBmp(folder + low_bright_file)
y_low, x_low = make_hist(low_bright[..., 0])
plt.bar(x_low, y_low)
plt.savefig(folder + 'original_low_hist.png')
plt.clf()
dot_a = (40, 65)
dot_b = (210, 180)
f = BmpFilters.two_dot_func(dot_a, dot_b)
r = range(0, 255)
y_grad = list(f(r))
x = list(r)
plt.plot(x, y_grad)
plt.savefig(folder + 'grad.png')
plt.clf()
low_fnc = list(map(lambda x: x * low_coeff, list(range(0, 255))))
x = list(r)
axes = plt.gca()
axes.set_xlim([0, 260])
axes.set_ylim([0, 260])
plt.plot(x, low_fnc)
plt.savefig(folder + 'low fnc.png')
plt.clf()
high_fnc = np.array(
list(map(lambda x: x * high_coeff, list(range(0, 255)))))
x = list(r)
np.putmask(high_fnc, high_fnc > 255, 255)
axes = plt.gca()
axes.set_xlim([0, 260])
axes.set_ylim([0, 260])
plt.plot(x, high_fnc)
plt.savefig(folder + 'high fnc.png')
plt.clf()
high_grad = f(np.array(high_fnc))
x = list(r)
np.putmask(high_grad, high_grad > 255, 255)
axes = plt.gca()
axes.set_xlim([0, 260])
axes.set_ylim([0, 260])
plt.plot(x, high_grad)
plt.savefig(folder + 'high fnc grad.png')
plt.clf()
low_grad = f(np.array(low_fnc))
x = list(r)
np.putmask(low_grad, low_grad > 255, 255)
axes = plt.gca()
axes.set_xlim([0, 260])
axes.set_ylim([0, 260])
plt.plot(x, low_grad)
plt.savefig(folder + 'low fnc grad.png')
plt.clf()
low_bright_grad = BmpFilters.BmpFilters.two_dot(low_bright, f)
y, x = make_hist(low_bright_grad[..., 0])
plt.bar(x, y)
plt.savefig(folder + 'low_grad_hist.png')
plt.clf()
high_bright_grad = BmpFilters.BmpFilters.two_dot(high_bright, f)
y, x = make_hist(high_bright_grad[..., 0])
plt.bar(x, y)
plt.savefig(folder + 'high_grad_hist.png')
plt.clf()
bmp.writeBmp(folder + 'low grad.bmp', header, low_bright_grad)
bmp.writeBmp(folder + 'high grad.bmp', header, high_bright_grad)
import BmpFilters
import bmp
orig_filename = 'lena.bmp'
median_folder = 'median'
original_file_y_cb_cr = 'lena y_cb_cr.bmp'
gauss_noise_filename = "gauss/gauss d =100.bmp"
moving_average_folder = "moving average"
gauss_noise_psnr_filename = "gauss/psnr.png"
impulse_noise_filename = "impulse/impulse.bmp"
gauss_filter_filename = "gauss filter"
laplas_folder = "laplas"
impule_noise_filename = 'impulse pa = 0.1 pb = 0.2.png'
_, orig_pic = bmp.readBmp(orig_filename)
def avg():
file_name = gauss_noise_filename
header, rgb = bmp.readBmp(file_name)
psnr = []
rng = range(1, 25)
y_cb_cr = bmp.convertYCbCr(rgb)
orinal_y_cb_cr = bmp.convertYCbCr(orig_pic)
for i in rng:
res = BmpFilters.BmpFilters.avg(y_cb_cr, i)
psnr.append(bmp.psnr(orinal_y_cb_cr[..., 0], res[..., 0]))
res[..., 1] = res[..., 0]
res[..., 2] = res[..., 0]
entropy_b = -np.sum(b * np.ma.log2(b).filled(0))
entropy_g = -np.sum(g * np.ma.log2(g).filled(0))
entropy_r = -np.sum(r * np.ma.log2(r).filled(0))
entropy_cb = -np.sum(cb * np.ma.log2(cb).filled(0))
entropy_cr = -np.sum(cr * np.ma.log2(cr).filled(0))
entropy_y = -np.sum(y * np.ma.log2(y).filled(0))
print("\n \n \n")
print(f"entropy type {func_type} of b ", entropy_b)
print(f"entropy type {func_type} of g ", entropy_g)
print(f"entropy type {func_type} of r ", entropy_r)
print(f"entropy type {func_type} of cb ", entropy_cb)
print(f"entropy type {func_type} of cr ", entropy_cr)
print(f"entropy type {func_type} of y ", entropy_y)
header, rgb = bmp.readBmp(in_file)
a = rgb[...][...][0] - rgb[...][...][1]
print(np.max(a))
sys.exit(0)
y_cb_cr = bmp.convertYCbCr(rgb)
cbDecimated = bmp.decimationByDeletingEven(y_cb_cr[..., 1], 4)
crDecimated = bmp.decimationByDeletingEven(y_cb_cr[..., 2], 4)
cbRecovered = bmp.recoverDecimationByDeletingEven(cbDecimated, 4)
crRecovered = bmp.recoverDecimationByDeletingEven(crDecimated, 4)
y_cb_cr[..., 1] = cbRecovered
y_cb_cr[..., 2] = crRecovered
rgb2 = bmp.inverseConvertYCbCr(y_cb_cr)
bmp.writeBmp(res_file_path + "testRec.bmp", header, rgb2)
print(bmp.psnr(rgb[..., 0], rgb2[..., 0]))
print(bmp.psnr(rgb[..., 1], rgb2[..., 1]))