def test_12bit_error():
# 1024x1024 のRampパターン作成
inc12 = np.arange(4096)
img_org = np.dstack([inc12, inc12, inc12])
img_org = img_org * np.ones((2160, 1, 3)) # V方向に拡張
img_normalized = img_org / np.max(img_org)
# 保存
fname = 'inc12.dpx'
attr_dpx = {"oiio:BitsPerSample": 12}
save_img_using_oiio(img_normalized, fname,
out_img_type_desc=oiio.UINT16, attr=attr_dpx)
# 読み込み
img_load, attr = load_img_using_oiio(fname)
img_load_12bit = np.uint16(np.round(normalize_by_dtype(img_load) * 4095))
# とりあえずプロット
ax1 = pu.plot_1_graph()
ax1.plot(img_load_12bit[0, :, 0])
plt.show()
# オリジナルデータとの差分確認
diff = np.sum(np.abs(img_org - img_load_12bit))
print(diff)
# 隣接ピクセルとの差分確認
line_data = img_load_12bit[0, :, 0]
diff = line_data[1:] - line_data[:-1]
print(np.sum(diff != 1))
def quadratic_bezier_curve(t, p0, p1, p2, samples=1024):
# x = ((1 - t) ** 2) * p0[0] + 2 * (1 - t) * t * p1[0]\
# + (t ** 2) * p2[0]
# y = ((1 - t) ** 2) * p0[1] + 2 * (1 - t) * t * p1[1]\
# + (t ** 2) * p2[1]
x = ((1 - t) ** 2) * p0[0] + 2 * (1 - t) * t * p1[0]\
+ (t ** 2) * p2[0]
y = ((1 - t) ** 2) * p0[1] + 2 * (1 - t) * t * p1[1]\
+ (t ** 2) * p2[1]
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(10, 8),
graph_title="Title",
graph_title_size=None,
xlabel="X Axis Label", ylabel="Y Axis Label",
axis_label_size=None,
legend_size=17,
xlim=None,
ylim=None,
xtick=None,
ytick=None,
xtick_size=None, ytick_size=None,
linewidth=3,
minor_xtick_num=None,
minor_ytick_num=None)
ax1.plot(x, y, label='aaa')
plt.legend(loc='upper left')
plt.show()
def plot_rrt():
fname = "./aces_1.0.3/luts/Dolby_PQ_4000_nits_Shaper.RRT.P3-D60_ST2084__4000_nits_.spi3d"
title = os.path.basename(fname)
x = tylut.load_3dlut_spi_format(fname)[0]
print(x.shape)
grid_num = 65
x2 = np.linspace(0, 1, grid_num)
y = []
for idx in range(grid_num):
line_idx = (grid_num ** 2) * idx + (grid_num ** 1) * idx\
+ idx
y.append(x[line_idx, 0])
y = np.array(y)
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(10, 8),
graph_title=title,
graph_title_size=None,
xlabel="Input",
ylabel="Output",
axis_label_size=None,
legend_size=17,
xlim=None,
ylim=(0, 1),
xtick=None,
ytick=None,
xtick_size=None,
ytick_size=None,
linewidth=3,
minor_xtick_num=None,
minor_ytick_num=None)
ax1.plot(x2, y)
plt.show()
def plot_n_log_stops():
x_max = tf.n_log_decoding(1.0)
x = get_log_scale_x(sample_num=64, x_max=x_max, stops=20)
gray18_linear_light = 0.20
y = tf.n_log_encoding(x) * 1023
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(10, 8),
graph_title="N-Log Characteristics",
graph_title_size=None,
xlabel="Input linear light stops",
ylabel="10bit code value",
axis_label_size=None,
legend_size=17,
xlim=[-8, 8],
ylim=[0, 1024],
xtick=[x for x in range(-8, 9)],
ytick=[x * 128 for x in range(8)],
xtick_size=None,
ytick_size=None,
linewidth=2)
ax1.plot(np.log2(x / gray18_linear_light), y, '-o', label="N-Log OETF")
plt.legend(loc='upper left')
plt.show()
def plot_rgb_histgram(r_data, g_data, b_data, title=None):
"""
RGB3色のヒストグラムを作成する
"""
plot_range = [-3, 3]
three_color_width = 0.7
each_width = three_color_width / 3
# データ生成
r = make_histogram_data(r_data, plot_range)
g = make_histogram_data(g_data, plot_range)
b = make_histogram_data(b_data, plot_range)
# plot
xtick = [x for x in range(plot_range[0], plot_range[1] + 1)]
ax1 = pu.plot_1_graph(graph_title=title,
graph_title_size=22,
xlabel="Difference",
ylabel="Frequency",
xtick=xtick,
grid=False)
range_r = np.arange(plot_range[0], plot_range[1] + 1) - each_width
range_g = np.arange(plot_range[0], plot_range[1] + 1)
range_b = np.arange(plot_range[0], plot_range[1] + 1) + each_width
ax1.bar(range_r, r[0], color=R_BAR_COLOR, label="Red", width=each_width)
ax1.bar(range_g, g[0], color=G_BAR_COLOR, label="Green", width=each_width)
ax1.bar(range_b, b[0], color=B_BAR_COLOR, label="Blue", width=each_width)
ax1.set_yscale("log", nonposy="clip")
plt.legend(loc='upper left')
fname = "figures/" + title + "three.png"
plt.savefig(fname, bbox_inches='tight', pad_inches=0.1)
plt.show()
def plot_log_stops():
# oetf_list = [tf.LOGC, tf.SLOG3, tf.DLOG, tf.FLOG, tf.NLOG]
oetf_list = [tf.LOGC, tf.SLOG3_REF]
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(16, 10),
graph_title="Characteristics of the camera log",
graph_title_size=None,
xlabel="Exposure [stops].",
ylabel="10bit code value",
axis_label_size=None,
legend_size=19,
xlim=[-8, 8],
ylim=[0, 1024],
xtick=[x for x in range(-8, 9)],
ytick=[x * 128 for x in range(9)],
xtick_size=None,
ytick_size=None,
linewidth=3)
# centerd_spins(ax1)
for oetf_name in oetf_list:
x_max = tf.MAX_VALUE[oetf_name]
x = get_log_scale_x(sample_num=64, x_max=x_max, stops=20)
x2 = x / 0.20 if oetf_name == tf.SLOG3 else x / 0.18
y = tf.oetf(x / x_max, oetf_name) * 1023
ax1.plot(np.log2(x2), y, '-o', label=oetf_name)
# Plot A-Log
alog = ALOG_DATA
x = get_log_scale_x(sample_num=64, x_max=ALOG_MAX, stops=20)
x2 = x / 0.18
ax1.plot(np.log2(x2), alog, '-o', label="Astrodesign A-Log??")
plt.legend(loc='upper left')
plt.savefig('camera_logs.png', bbox_inches='tight', pad_inches=0.1)
plt.show()
def plot_from_white_to_primary_rgb_value_with_bar(
primary_color='green', step=5, name=cs.BT709, oetf_name=tf.GAMMA24):
rgb = get_from_white_to_primary_rgb_value(
'green', step=5, name=cs.BT709, oetf_name=tf.GAMMA24)
normalize_coef = np.sum(rgb, axis=-1).reshape((rgb.shape[0], 1))
rgb_normalized = rgb / normalize_coef * 100
x_caption = ["No.{}".format(x) for x in range(rgb.shape[0])]
x_val = np.arange(rgb.shape[0])
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(9, 9),
graph_title="色とRGBの割合の関係",
xlabel=None,
ylabel="Percentage of RGB values [%]",
ylim=(0, 105))
for no_idx in range(rgb.shape[0]):
bottom = 0
for c_idx in range(3):
x = no_idx
y = rgb_normalized[no_idx][c_idx]
color = RGB_COLOUR_LIST[c_idx]
ax1.bar(x, y, bottom=bottom, width=0.7, color=color)
bottom += y
plt.xticks(x_val, x_caption)
plt.savefig('stacked_graph.png', bbox_inches='tight', pad_inches=0.1)
plt.show()
def plot_d65():
d65_file = "./src_data/d65_CIE_S_014-2.csv"
d65 = np.loadtxt(d65_file, delimiter=',').T
wl = d65[0].flatten()
d65_spd = d65[2].flatten()
d65_spd_sprague = make_day_light_by_calculation(
temperature=6504.0, interpolater=SpragueInterpolator)
d65_spd_linear = make_day_light_by_calculation(
temperature=6504.0, interpolater=LinearInterpolator)
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(10, 8),
linewidth=5,
graph_title="CIE Illuminant D65",
xlabel="wavelength [nm]",
ylabel="rerative spectral power distributions")
ax1.plot(wl, d65_spd, '-', color=R_PLOT_COLOR, label='D65 CIE S 014-2')
ax1.plot(wl,
d65_spd_linear[1, ...],
'--',
lw=2,
color=G_PLOT_COLOR,
label='D65 Calc with LinearInterpolation')
ax1.plot(wl,
d65_spd_sprague[1, ...],
'-',
lw=2,
color=B_PLOT_COLOR,
label='D65 Calc with SpragueInterpolation')
plt.legend(loc='upper left')
plt.show()
def compare_1nm_value_and_target_value(spd, org_cmfs):
"""
各種変換方式と、公式の1nmのCMFSの誤差を比較
"""
# intp_methods = ["linear", "sprague", "spline"]
intp_methods = ["sprague", "spline"]
spds = [
interpolate_5nm_cmfs_data(spd, intp_method)
for intp_method in intp_methods
]
wl = spds[0].wavelengths
diffs = [org_cmfs - spds[idx].values for idx in range(len(intp_methods))]
error_rates = np.array([diff / org_cmfs for diff in diffs])
error_rates[np.isinf(error_rates)] = 0
error_rates[np.isnan(error_rates)] = 0
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(10, 8),
linewidth=3,
graph_title="Comparison of interpolation methods.",
xlabel="Wavelength [nm]",
ylabel='Error rate of y [%]',
xlim=[490, 510],
ylim=[-0.1, 0.1])
for idx in range(len(intp_methods)):
y = error_rates[idx, :, 2] * 100
ax1.plot(wl, y, '-o', label=intp_methods[idx])
plt.legend(loc='lower right')
plt.show()
def plot_rgb_stacked_bar_graph(data):
"""
Examples
--------
>>> data = {"ACES AP0": np.array([[100, 0, 0], [0, 100, 0], [0, 0, 100]]),
"BT.709": np.array([[70, 20, 10], [10, 70, 20], [20, 10, 70]]),
"BT.2020": np.array([[85, 10, 5], [5, 85, 10], [10, 5, 85]])}
>>> plot_rgb_stacked_bar_graph(data)
"""
x_val = np.arange(len(data)) + 1
x_offset = [-0.25, 0.0, 0.25]
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(12, 9),
graph_title="Color Space Conversion to ACES AP0",
xlabel="Source Gamut",
ylabel="Percentage of RGB values [%]",
ylim=(0, 105))
for gg, gamut in enumerate(data): # 色域のループ
value = data[gamut]
value[value < 0] = 0
value[value > 1] = 1
normalize_val = np.sum(value, axis=-1) / 100
value = value / normalize_val.reshape(value.shape[0], 1)
for ii in range(3): # 原色が3種類ある、のループ
x = x_val[gg] + x_offset[ii]
bottom = 0
for jj in range(3): # R, G, B の各要素のループ
y = value[ii][jj]
color = RGB_COLOUR_LIST[jj]
ax1.bar(x, y, bottom=bottom, width=0.2, color=color)
bottom += y
plt.xticks(x_val, list(data.keys()))
plt.savefig('stacked_ap0.png', bbox_inches='tight', pad_inches=0.1)
plt.show()
def plot_shaper():
fname = "./aces_1.0.3/luts/Dolby_PQ_1000_nits_Shaper_to_linear.spi1d"
title = os.path.basename(fname)
x = tylut.load_1dlut_spi_format(fname)[..., 0]
sample_num = len(x)
y = np.linspace(0, 1, sample_num)
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(10, 8),
graph_title=title,
graph_title_size=None,
xlabel="Linear (log scale)",
ylabel="Output",
axis_label_size=None,
legend_size=17,
xlim=None,
ylim=None,
xtick=None,
ytick=None,
xtick_size=None,
ytick_size=None,
linewidth=3,
minor_xtick_num=None,
minor_ytick_num=None)
ax1.plot(np.log2(x / 0.18), y)
plt.show()
def canon_log3_eotf(x, plot=False):
"""
Defines the *Canon Log 3* log decoding curve / electro-optical transfer
function. This function is based on log_decoding_CanonLog3 in Colour.
Parameters
----------
x : numeric or ndarray
normalized code value. Interval is [0:1] .
Returns
-------
numeric or ndarray
Linear data y.
Notes
-----
- Input *x* code value is converted to *IRE* as follows:
`IRE = (CV - 64) / (940 - 64)`.
Examples
--------
>>> x = np.linspace(0, 1, 1024)
>>> y = canon_log3_eotf(x, plot=True)
"""
# convert x to the 10bit code value
code_value = x * 1023
# convert code_value to ire value
ire = (code_value - 64) / (940 - 64)
y = np.select((ire < 0.04076162, ire <= 0.105357102, ire > 0.105357102),
(-(10**((0.069886632 - ire) / 0.42889912) - 1) / 14.98325,
(ire - 0.073059361) / 2.3069815,
(10**((ire - 0.069886632) / 0.42889912) - 1) / 14.98325))
if plot:
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(10, 8),
graph_title="Title",
graph_title_size=None,
xlabel="X Axis Label",
ylabel="Y Axis Label",
axis_label_size=None,
legend_size=17,
xlim=None,
ylim=None,
xtick=None,
ytick=None,
xtick_size=None,
ytick_size=None,
linewidth=3)
ax1.plot(x, y, label="Canon Log3 EOTF")
plt.legend(loc='upper left')
plt.show()
return y
def canon_log2_eotf(x, plot=False):
"""
Defines the *Canon Log 2* log decoding curve / electro-optical transfer
function. This function is based on log_decoding_CanonLog2 in Colour.
Parameters
----------
x : numeric or ndarray
normalized code value. Interval is [0:1] .
Returns
-------
numeric or ndarray
Linear data y.
Notes
-----
- Input *x* code value is converted to *IRE* as follows:
`IRE = (CV - 64) / (940 - 64)`.
Examples
--------
>>> x = np.linspace(0, 1, 1024)
>>> y = canon_log2_eotf(x, plot=True)
"""
# convert x to the 10bit code value
code_value = x * 1023
# convert code_value to ire value
clog2_ire = (code_value - 64) / (940 - 64)
y = np.where(
clog2_ire < 0.035388128,
-(10**((0.035388128 - clog2_ire) / 0.281863093) - 1) / 87.09937546,
(10**((clog2_ire - 0.035388128) / 0.281863093) - 1) / 87.09937546)
if plot:
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(10, 8),
graph_title="Title",
graph_title_size=None,
xlabel="X Axis Label",
ylabel="Y Axis Label",
axis_label_size=None,
legend_size=17,
xlim=None,
ylim=None,
xtick=None,
ytick=None,
xtick_size=None,
ytick_size=None,
linewidth=3)
ax1.plot(x, y, label="Canon Log2 EOTF")
plt.legend(loc='upper left')
plt.show()
return y
def canon_log3_oetf(x, plot=False):
"""
Defines the *Canon Log 3* log encoding curve / opto-electronic transfer
function. This function is based on log_encoding_CanonLog3 in Colour.
Parameters
----------
x : numeric or ndarray
Linear data.
Returns
-------
numeric or ndarray
Canon Log3 normalized code value(10bit).
Notes
-----
- output *y* code value is converted to *normalized code value*
as follows: `NCV = (IRE * (940 - 64) + 64 / 1023)` .
Examples
--------
>>> linear_max = canon_log3_eotf(1.0, plot=True)
>>> x = np.linspace(0, 1, 1024)
>>> y = canon_log3_eotf(x * linear_max, plot=True)
"""
clog3_ire = np.select(
(x < _log_decoding_CanonLog3(0.04076162),
x <= _log_decoding_CanonLog3(0.105357102),
x > _log_decoding_CanonLog3(0.105357102)),
(-(0.42889912 *
(np.log10(-x * 14.98325 + 1)) - 0.069886632), 2.3069815 * x +
0.073059361, 0.42889912 * np.log10(x * 14.98325 + 1) + 0.069886632))
code_value = (clog3_ire * (940 - 64) + 64) / 1023
if plot:
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(10, 8),
graph_title="Title",
graph_title_size=None,
xlabel="X Axis Label",
ylabel="Y Axis Label",
axis_label_size=None,
legend_size=17,
xlim=None,
ylim=None,
xtick=None,
ytick=None,
xtick_size=None,
ytick_size=None,
linewidth=3)
ax1.plot(x, code_value, label="Canon Log3 OETF")
plt.legend(loc='upper left')
plt.show()
return code_value
def plot_log_graph():
stops = 29
sample_num = 1024
# Canon Log1
max_val1 = canon_log_eotf(1.0, plot=False)
x = np.linspace(0, stops, sample_num)
logx1 = (1 / 2**(x)) * max_val1
clog1 = canon_log_oetf(logx1, plot=False)
# Canon Log2
max_val2 = canon_log2_eotf(1.0, plot=False)
x = np.linspace(0, stops, sample_num)
logx2 = (1 / 2**(x)) * max_val2
clog2 = canon_log2_oetf(logx2, plot=False)
# Canon Log3
max_val3 = canon_log3_eotf(1.0, plot=False)
x = np.linspace(0, stops, sample_num)
logx3 = (1 / 2**(x)) * max_val3
clog3 = canon_log3_oetf(logx3, plot=False)
# ACEScct
max_val_aces_cct = 65504
x = np.linspace(0, stops, sample_num)
logx_aces_cct = (1 / 2**(x)) * max_val_aces_cct
aces_cct = aces_cct_oetf(logx_aces_cct, plot=False)
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(16, 10),
graph_title="Characteristics of the OETF",
graph_title_size=None,
xlabel="Relative Stop from 18% Gray",
ylabel="10bit Code Value",
axis_label_size=None,
legend_size=17,
xlim=(-10, 10),
ylim=(0, 1024),
xtick=[x for x in range(-10, 11)],
ytick=[x * 64 for x in range(1024 // 64 + 1)],
xtick_size=None,
ytick_size=None,
linewidth=3)
x_axis = np.log2(logx1 / 0.20)
y_axis = clog1 * 1023
ax1.plot(x_axis, y_axis, 'r-', label="Hoge Log")
x_axis = np.log2(logx2 / 0.20)
y_axis = clog2 * 1023
ax1.plot(x_axis, y_axis, 'b-', label="Hoge Log2")
x_axis = np.log2(logx3 / 0.20)
y_axis = clog3 * 1023
ax1.plot(x_axis, y_axis, 'g-', label="Hoge Log3")
x_axis = np.log2(logx_aces_cct / aces_cct_oetf(0.18))
y_axis = aces_cct * 1023
ax1.plot(x_axis, y_axis, 'c-', label="ACEScct")
plt.legend(loc='upper left')
plt.show()
def get_l_focal(hue=45):
"""
hueから L_focal を得る
"""
# まずは L_cusp を求める
# ---------------------
lab_709, lab_2020, rgb = get_lab_edge(hue)
chroma_709 = get_chroma(lab_709)
chroma_2020 = get_chroma(lab_2020)
bt709_cusp_idx = np.argmax(chroma_709)
bt2020_cusp_idx = np.argmax(chroma_2020)
bt709_point = sympy.Point(chroma_709[bt709_cusp_idx],
lab_709[bt709_cusp_idx, 0])
bt2020_point = sympy.Point(chroma_2020[bt2020_cusp_idx],
lab_2020[bt2020_cusp_idx, 0])
chroma_line = sympy.Line(bt709_point, bt2020_point)
lightness_line = sympy.Line(sympy.Point(0, 0), sympy.Point(0, 100))
intersection = sympy.intersection(chroma_line, lightness_line)[0].evalf()
l_cusp = np.array(intersection)
# BT.2407 に従って補正
# ---------------------
# plot
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(10, 8),
graph_title=None,
graph_title_size=None,
xlabel="Chroma",
ylabel="Lightness",
axis_label_size=None,
legend_size=17,
xlim=[0, 220],
ylim=[0, 100],
xtick=None,
ytick=None,
xtick_size=None, ytick_size=None,
linewidth=3)
ax1.plot(chroma_709, lab_709[..., 0], c="#808080", label='BT.709')
ax1.plot(chroma_2020, lab_2020[..., 0], c="#000000", label='BT.2020')
ax1.plot(chroma_709[bt709_cusp_idx], lab_709[bt709_cusp_idx, 0], 'or',
markersize=10, alpha=0.5)
ax1.plot(chroma_2020[bt2020_cusp_idx], lab_2020[bt2020_cusp_idx, 0], 'or',
markersize=10, alpha=0.5)
ax1.plot(l_cusp[0], l_cusp[1], 'ok', markersize=10, alpha=0.5)
# annotation
ax1.annotate(r'L^*_{cusp}', xy=(l_cusp[0], l_cusp[1]),
xytext=(l_cusp[0] + 10, l_cusp[1] + 10),
arrowprops=dict(facecolor='black', shrink=0.1))
ax1.plot([chroma_2020[bt2020_cusp_idx], l_cusp[0]],
[lab_2020[bt2020_cusp_idx, 0], l_cusp[1]], '--k', alpha=0.3)
plt.legend(loc='upper right')
plt.show()
def canon_log2_oetf(x, plot=False):
"""
Defines the *Canon Log 2* log encoding curve / opto-electronic transfer
function. This function is based on log_encoding_CanonLog2 in Colour.
Parameters
----------
x : numeric or ndarray
Linear data.
Returns
-------
numeric or ndarray
Canon Log2 normalized code value(10bit).
Notes
-----
- output *y* code value is converted to *normalized code value*
as follows: `NCV = (IRE * (940 - 64) + 64 / 1023)` .
Examples
--------
>>> linear_max = canon_log2_eotf(1.0, plot=True)
>>> x = np.linspace(0, 1, 1024)
>>> y = canon_log2_eotf(x * linear_max, plot=True)
"""
clog2_ire = np.where(
x < _log_decoding_CanonLog2(0.035388128),
-(0.281863093 * (np.log10(-x * 87.09937546 + 1)) - 0.035388128),
0.281863093 * np.log10(x * 87.09937546 + 1) + 0.035388128)
code_value = (clog2_ire * (940 - 64) + 64) / 1023
if plot:
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(10, 8),
graph_title="Title",
graph_title_size=None,
xlabel="X Axis Label",
ylabel="Y Axis Label",
axis_label_size=None,
legend_size=17,
xlim=None,
ylim=None,
xtick=None,
ytick=None,
xtick_size=None,
ytick_size=None,
linewidth=3)
ax1.plot(x, code_value, label="Canon Log2 OETF")
plt.legend(loc='upper left')
plt.show()
return code_value
def plot_chromaticity_diagram(gamut, data, title=None):
xyY = ryr.rgb_to_xyY(data, gamut)
gamut_xy, _ = tpg.get_primaries(gamut)
cmf_xy = tpg._get_cmfs_xy()
rate = 1.0
ax1 = pu.plot_1_graph(fontsize=20 * rate,
figsize=(8 * rate, 9 * rate),
graph_title=title,
graph_title_size=16,
xlabel=None,
ylabel=None,
axis_label_size=None,
legend_size=18 * rate,
xlim=(0, 0.8),
ylim=(0, 0.9),
xtick=[x * 0.1 for x in range(9)],
ytick=[x * 0.1 for x in range(10)],
xtick_size=17 * rate,
ytick_size=17 * rate,
linewidth=4 * rate,
minor_xtick_num=2,
minor_ytick_num=2)
color = data.reshape((data.shape[0] * data.shape[1], data.shape[2]))
ax1.plot(cmf_xy[..., 0], cmf_xy[..., 1], '-k', lw=3.5 * rate, label=None)
ax1.plot((cmf_xy[-1, 0], cmf_xy[0, 0]), (cmf_xy[-1, 1], cmf_xy[0, 1]),
'-k',
lw=2.5 * rate,
label=None)
ax1.patch.set_facecolor("#F2F2F2")
ax1.plot(gamut_xy[..., 0],
gamut_xy[..., 1],
c=K_BAR_COLOR,
label="BT.709",
lw=3 * rate)
ax1.scatter(xyY[..., 0],
xyY[..., 1],
s=2 * rate,
marker='o',
c=color,
edgecolors=None,
linewidth=1 * rate,
zorder=100)
ax1.scatter(np.array([0.3127]),
np.array([0.3290]),
s=150 * rate,
marker='x',
c="#000000",
edgecolors=None,
linewidth=2.5 * rate,
zorder=101,
label="D65")
plt.legend(loc='upper right')
file_name = './figures/xy_chromaticity_{}.png'.format(title)
plt.savefig(file_name, bbox_inches='tight')
def aces_cct_eotf(x, plot=False):
"""
Defines the *ACEScct* log decoding curve / electro-optical transfer
function.
reference : http://j.mp/S-2016-001_
Parameters
----------
x : numeric or ndarray
normalized code value. Interval is [0:1] .
Returns
-------
numeric or ndarray
Linear data y.
Notes
-----
None
Examples
--------
>>> max_val = aces_cct_oetf(65504)
>>> x = np.linspace(0, 1, 1024)
>>> y = aces_cct_eotf(x * max_val, plot=True)
"""
y = np.select(
(x <= 0.155251141552511, x < (np.log2(65504) + 9.72) / 17.52, x >=
(np.log2(65504) + 9.72) / 17.52),
((x - 0.0729055341958355) / 10.5402377416545, 2
**(x * 17.52 - 9.72), 65504))
if plot:
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(10, 8),
graph_title="Title",
graph_title_size=None,
xlabel="X Axis Label",
ylabel="Y Axis Label",
axis_label_size=None,
legend_size=17,
xlim=None,
ylim=None,
xtick=None,
ytick=None,
xtick_size=None,
ytick_size=None,
linewidth=3)
ax1.plot(x, y, label="ACEScct EOTF")
plt.legend(loc='upper left')
plt.show()
return y
def _plot_comparison_between_1000nits_and_108nits_primaries(
x, color_idx, cs_name, min_value, max_value,
ot_108_img, ot_1000_img,
x_min_exposure, x_max_exposure):
title_base = "OutputTrasform comparison {}_{}"
title_str = title_base.format(cs_name, COLOR_NAME_LIST[color_idx])
ax1 = pu.plot_1_graph(
fontsize=20,
figsize=(16, 10),
graph_title=title_str,
graph_title_size=None,
xlabel="Linear (center is 18% gray)",
ylabel='Luminance [nits]',
axis_label_size=None,
legend_size=20,
xlim=None,
ylim=(min_value, max_value),
xtick=None,
ytick=None,
xtick_size=16, ytick_size=None,
linewidth=3,
minor_xtick_num=None,
minor_ytick_num=None)
ax1.set_xscale('log', basex=2.0)
ax1.set_yscale('log', basey=10.0)
y_600nits = np.ones_like(x) * 600
x_val_num = x_max_exposure - x_min_exposure + 1
x_val = [0.18 * (2 ** (x + x_min_exposure))
for x in range(x_val_num) if x % 2 == 0]
x_caption = [r"$0.18 \times 2^{{{}}}$".format(x + x_min_exposure)
for x in range(x_val_num) if x % 2 == 0]
ax1.plot(x, ot_1000_img[color_idx, :, 0], '-', color=RGB_COLOUR_LIST[0],
label="Output Transform BT2020 1000nits")
ax1.plot(x, ot_1000_img[color_idx, :, 1], '-', color=RGB_COLOUR_LIST[1],
label="Output Transform BT2020 1000nits")
ax1.plot(x, ot_1000_img[color_idx, :, 2], '-', color=RGB_COLOUR_LIST[2],
label="Output Transform BT2020 1000nits")
ax1.plot(x, ot_108_img[color_idx, :, 0], '--', color=RGB_COLOUR_LIST[0],
label="Output Transform P3D65 108nits")
ax1.plot(x, ot_108_img[color_idx, :, 1], '--', color=RGB_COLOUR_LIST[1],
label="Output Transform P3D65 108nits")
ax1.plot(x, ot_108_img[color_idx, :, 2], '--', color=RGB_COLOUR_LIST[2],
label="Output Transform P3D65 108nits")
ax1.plot(x, y_600nits, '-', color="#606060",
label="600nits", lw=1)
plt.xticks(x_val, x_caption)
plt.legend(loc='upper left')
img_name_base = "comparison_108_vs_1000_primaries_{}_{}.png"
img_name = img_name_base.format(cs_name, COLOR_NAME_LIST[color_idx])
plt.savefig(img_name, bbox_inches='tight', pad_inches=0.1)
plt.show()
def tonemap_2dim_bezier(x, top_param, bottom_param, plot=False):
"""
2次ベジェ曲線でトーンマップする
Parameters
----------
x : array_like
target data.
top_param : dictionary
tonemapping parameters for top area.
bottom_param : dictionary
tonemapping parameters for bottom area.
plot : boolean
whether plotting or not.
Note
----
An example of ```top_param``` is bellow.
{'x0': 0.5, 'y0': 0.5,
'x1': 0.7, 'y1': 0.7,
'x2': 1.0, 'y2': 0.7}
An example of ```bottom_param``` is bellow.
{'x0': 0.0, 'y0': 0.1,
'x1': 0.1, 'y1': 0.1,
'x2': 0.3, 'y2': 0.3}
"""
temp = tonemap_2dim_bezier_bottom(x, plot=False, **bottom_param)
y = tonemap_2dim_bezier_top(temp, plot=False, **top_param)
if plot:
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(10, 8),
graph_title="Title",
graph_title_size=None,
xlabel="X Axis Label",
ylabel="Y Axis Label",
axis_label_size=None,
legend_size=17,
xlim=None,
ylim=None,
xtick=None,
ytick=None,
xtick_size=None,
ytick_size=None,
linewidth=3)
ax1.plot(x, y, label='bezier')
plt.legend(loc='upper left')
plt.show()
return y
def plot_log_basic(name=tf.FLOG, reflection=False):
if name == tf.FLOG:
encode_func = tf.f_log_encoding
decode_func = tf.f_log_decoding
elif name == tf.NLOG:
encode_func = tf.n_log_encoding
decode_func = tf.n_log_decoding
elif name == tf.DLOG:
encode_func = tf.d_log_encoding
decode_func = tf.d_log_decoding
else:
raise ValueError("not supported log name")
x = np.linspace(0, 1, 1024)
y = decode_func(x, out_reflection=reflection)
ax1 = pu.plot_1_graph(linewidth=3)
ax1.plot(x, y, label=name + " EOTF")
plt.legend(loc='upper left')
plt.show()
x_max = decode_func(1.0, out_reflection=reflection)
x = np.linspace(0, 1, 1024) * x_max
y = encode_func(x, in_reflection=reflection)
ax1 = pu.plot_1_graph(linewidth=3)
ax1.plot(x, y, label=name + " OETF")
plt.legend(loc='upper left')
plt.show()
y2 = decode_func(y, out_reflection=reflection)
ax1 = pu.plot_1_graph(linewidth=3)
ax1.plot(x, y2, label="Linear")
plt.legend(loc='upper left')
plt.show()
print(np.max(np.abs(y2 - x)))
print(y2)
def plot_d65_and_calculated_d65():
"""
計算で出した D65 と エクセルに値が書かれていた D65 で
スペクトル分が違うのかプロットしてみる。
"""
t = np.array([6500.0])
wl_org, s_org = lit.get_d_illuminants_spectrum(t)
wl_calc, s_calc = lit.get_d65_spectrum()
s_org = s_org[0]
ax1 = pu.plot_1_graph()
ax1.plot(wl_org, s_org, '-o', label="org")
ax1.plot(wl_calc, s_calc, '-o', label="calc")
plt.legend(loc='upper left')
plt.show()
def plot_lab_leaf(sample_num=9):
"""
chroma-lightness の leafをプロットする。
RGBMYCの6パターンでプロットする。
Parameters
----------
sample_num : int
sample number for each data.
"""
rgb = rgbmyc_data_for_lab(sample_num)
lab_709 = rgb_to_lab_d65(rgb=rgb, name="ITU-R BT.709")
lab_2020 = rgb_to_lab_d65(rgb=rgb, name="ITU-R BT.2020")
l_709 = lab_709[..., 0]
l_2020 = lab_2020[..., 0]
c_709 = get_chroma(lab_709)
c_2020 = get_chroma(lab_2020)
h_709 = get_hue(lab_709)
h_2020 = get_hue(lab_2020)
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(10, 8),
graph_title="Title",
graph_title_size=None,
xlabel="Chroma",
ylabel="Lightness",
axis_label_size=None,
legend_size=17,
xlim=None,
ylim=None,
xtick=None,
ytick=None,
xtick_size=None, ytick_size=None,
linewidth=3)
ax1.plot(c_709[0, ...], l_709[0, ...], '-o',
c="#FF8080", label="BT.709" + " " + str(h_709[0, ...]))
ax1.plot(c_2020[0, ...], l_2020[0, ...], '-o',
c="#FF4040", label="BT.2020" + " " + str(h_2020[0, ...]))
plt.legend(loc='upper right')
plt.show()
def plot_cmfs():
cmfs_file = "./src_data/CMFs_CIE_S_014-1.csv"
cmfs = np.loadtxt(cmfs_file, delimiter=',').T
wl = cmfs[0].flatten()
x = cmfs[1].flatten()
y = cmfs[2].flatten()
z = cmfs[3].flatten()
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(10, 8),
graph_title="CIE 1931 2-deg, XYZ CMFs",
xlabel="wavelength [nm]",
ylabel="tristimulus values")
ax1.plot(wl, x, '-r', label='x')
ax1.plot(wl, y, '-g', label='y')
ax1.plot(wl, z, '-b', label='z')
plt.legend(loc='upper left')
plt.show()
def get_bt2100_hlg_curve(x, debug=False):
"""
参考：ITU-R Recommendation BT.2100-0
"""
a = 0.17883277
b = 0.28466892
c = 0.55991073
under = (x <= 0.5) * 4 * (x**2)
over = (x > 0.5) * (np.exp((x - c) / a) + b)
y = (under + over) / 12.0
if debug:
ax1 = pu.plot_1_graph()
ax1.plot(x, y)
plt.show()
return y
def plot_gray_scale_exr_0_1_3dlut_linear():
sample_num = 1024
x_m10_10 = np.linspace(-10, 10, sample_num)
img = read_exr_data("./exp_min_0_max_1_out_3dlut_linear.exr")
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(16, 10),
graph_title="3DLUTの範囲:0～1。allocation: uniform",
xlabel="input value",
ylabel="output value",
xlim=[-0.5, 1.5],
linewitdh=3)
# ax1.plot(x_0_1, img[0, 0:sample_num, 0].flatten(), label="0～1")
# ax1.plot(x_m1_2, img[1, 0:sample_num, 0].flatten(), label="-1～2")
ax1.plot(x_m10_10, img[2, 0:sample_num, 0].flatten(), '-o', label="-10～10")
plt.legend(loc='upper left')
fname = './figures/graph_0_1_3dlut_linear.png'
plt.savefig(fname, bbox_inches='tight', pad_inches=0.1)
plt.show()
def plot_diff(a, b):
"""
plot the difference between a and b.
Parameters
----------
a : ndarray
data 1
b : ndarray
data 2
Returns
-------
none.
Notes
-----
none.
Examples
--------
>>> a = np.linspace(0, 1, 1024)
>>> b = np.arange(1024) / 1023
>>> plot_diff(a, b)
"""
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(10, 8),
graph_title="Diff",
graph_title_size=None,
xlabel="Code Value",
ylabel="Difference (Linear)",
axis_label_size=None,
legend_size=17,
xlim=(0, 1024),
ylim=None,
xtick=[x * 128 for x in range(0, 1024 // 128 + 1)],
ytick=None,
xtick_size=None,
ytick_size=None,
linewidth=3)
ax1.plot((a - b), '-o', label="diff")
plt.legend(loc='upper left')
plt.show()
def d_illuminant_interpolation(plot=False):
"""
# 概要
D Illuminant の分光特性は 10nm 刻みなので、
それを 5nm 刻みに線形補完する
"""
# D Illuminant 算出 (10nm精度)
# -----------------------------
t = np.array([5000], dtype=np.float64)
wl, s = lit.get_d_illuminants_spectrum(t)
# Interpolation
# -----------------------------
# f_quadratic = interpolate.interp1d(wl, s, kind='quadratic')
f_cubic = interpolate.interp1d(wl, s, kind='cubic')
wl_new = np.arange(380, 785, 5, dtype=np.uint16)
# s_quadratic = f_quadratic(wl_new)
s_cubic = f_cubic(wl_new)
if plot:
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(10, 8),
graph_title="cubic interpolation",
graph_title_size=None,
xlabel="Wavelength [nm]",
ylabel="Intensity",
axis_label_size=None,
legend_size=17,
xlim=None,
ylim=None,
xtick=None,
ytick=None,
xtick_size=None,
ytick_size=None,
linewidth=None)
ax1.plot(wl, s, 'ro', linewidth=5, label="original")
# ax1.plot(wl_new, s_quadratic, 'g-x', linewidth=2, label="quadratic")
ax1.plot(wl_new, s_cubic, 'b-', linewidth=2, label="cubic")
plt.legend(loc='lower right')
plt.show()
return wl_new, s_cubic
def plot_ab_pane_of_lab(sample_num):
"""
L*a*b*空間の a*b* 平面をプロットする。
Hueの考え方がxyY空間と同じか確認するため。
データは RGBMYC の6種類（RGBベース）。
Parameters
----------
sample_num : int
sample number for each data.
"""
rgb = rgbmyc_data_for_lab(sample_num)
lab_709 = rgb_to_lab_d65(rgb=rgb, name="ITU-R BT.709")
lab_2020 = rgb_to_lab_d65(rgb=rgb, name="ITU-R BT.2020")
color_709 = ["#FF8080", "#80FF80", "#8080FF",
"#FF80FF", "#FFFF80", "#80FFFF"]
color_2020 = ["#FF0000", "#00FF00", "#0000FF",
"#FF00FF", "#FFFF00", "#00FFFF"]
ax1 = pu.plot_1_graph(fontsize=20,
figsize=(10, 8),
graph_title="Title",
graph_title_size=None,
xlabel="a*",
ylabel="b*",
axis_label_size=None,
legend_size=17,
xlim=None,
ylim=None,
xtick=None,
ytick=None,
xtick_size=None, ytick_size=None,
linewidth=3)
for idx in range(rgb.shape[0]):
ax1.plot(lab_709[idx, :, 1], lab_709[idx, :, 2], '-o',
c=color_709[idx], label="BT.709_" + str(idx))
ax1.plot(lab_2020[idx, :, 1], lab_2020[idx, :, 2], '-o',
c=color_2020[idx], label="BT.2020_" + str(idx))
plt.legend(loc='lower left')
plt.show()