Skip to content

Commit

Permalink
Use methods already defined in colour instead of re-implementing them.
Browse files Browse the repository at this point in the history
  • Loading branch information
@MichaelMauderer
MichaelMauderer committed Jan 8, 2025
1 parent f63f390 commit 77c42f8
Showing 1 changed file with 71 additions and 87 deletions.
158 changes: 71 additions & 87 deletions colour/io/luts/clf.py
View file
Original file line number Diff line number Diff line change
Expand Up @@ -3,7 +3,6 @@
"""
import colour_clf_io as clf
import numpy as np
import numpy.lib.scimath as emath

from colour.algebra import (
table_interpolation_tetrahedral,
Expand All @@ -21,6 +20,14 @@
from colour_clf_io import ExponentStyle
from colour_clf_io.values import Channel

from colour.models.rgb.transfer_functions import (
exponent_function_basic,
exponent_function_monitor_curve,
logarithmic_function_basic,
logarithmic_function_camera,
logarithmic_function_quasilog,
)


def from_uint16_to_f16(array: npt.NDArray[np.uint16]) -> npt.NDArray[np.float16]:
values = list(map(int, array))
Expand Down Expand Up @@ -151,32 +158,46 @@ def apply_log_internal(value: NDArrayFloat, params, extra_args) -> NDArrayFloat:
style, in_bit_depth, out_bit_depth = extra_args
match style:
case clf.LogStyle.LOG_10:
return np.log10(np.maximum(value, FLT_MIN)) / out_bit_depth.scale_factor()
return (
logarithmic_function_basic(np.maximum(value, FLT_MIN), "log10")
/ out_bit_depth.scale_factor()
)
case clf.LogStyle.ANTI_LOG_10:
return np.power(10, value)
return logarithmic_function_basic(np.maximum(value, FLT_MIN), "antiLog10")
case clf.LogStyle.LOG_2:
return np.log2(np.maximum(value, FLT_MIN))
return logarithmic_function_basic(np.maximum(value, FLT_MIN), "log2")
case clf.LogStyle.ANTI_LOG_2:
return np.power(2, value)
return logarithmic_function_basic(np.maximum(value, FLT_MIN), "antiLog2")
case clf.LogStyle.LIN_TO_LOG:
log_side_slope = params.log_side_slope
lin_side_slope = params.lin_side_slope
log_side_offset = params.log_side_offset
lin_side_offset = params.lin_side_offset
base = params.base
log_side_value = emath.logn(
return logarithmic_function_quasilog(
value,
"linToLog",
base,
np.maximum(lin_side_slope * value + lin_side_offset, FLT_MIN),
log_side_slope,
lin_side_slope,
log_side_offset,
lin_side_offset,
)
return log_side_slope * log_side_value + log_side_offset
case clf.LogStyle.LOG_TO_LIN:
log_side_slope = params.log_side_slope
lin_side_slope = params.lin_side_slope
log_side_offset = params.log_side_offset
lin_side_offset = params.lin_side_offset
base = params.base
lin_side_value = base ** ((value - log_side_offset) / log_side_slope)
return (lin_side_value - lin_side_offset) / lin_side_slope
return logarithmic_function_quasilog(
value,
"logToLin",
base,
log_side_slope,
lin_side_slope,
log_side_offset,
lin_side_offset,
)
case clf.LogStyle.CAMERA_LIN_TO_LOG:
log_side_slope = params.log_side_slope
lin_side_slope = params.lin_side_slope
Expand All @@ -185,30 +206,16 @@ def apply_log_internal(value: NDArrayFloat, params, extra_args) -> NDArrayFloat:
lin_side_break = params.lin_side_break
base = params.base
linear_slope = params.linear_slope
if lin_side_slope is None:
linear_slope = (
log_side_slope
* lin_side_slope
/ (
(lin_side_slope * lin_side_break + lin_side_offset)
* np.log(base)
)
)
log_side_break = (
log_side_slope
* emath.logn(base, lin_side_slope * lin_side_break + lin_side_offset)
+ log_side_offset
)
linear_offset = log_side_break - linear_slope * lin_side_break
return np.where(
value < lin_side_break,
linear_slope * value + linear_offset,
log_side_slope
* emath.logn(
base,
np.maximum(lin_side_slope * value + lin_side_offset, FLT_MIN),
)
+ log_side_offset,
return logarithmic_function_camera(
value,
"cameraLinToLog",
base,
log_side_slope,
lin_side_slope,
log_side_offset,
lin_side_offset,
lin_side_break,
linear_slope,
)
case clf.LogStyle.CAMERA_LOG_TO_LIN:
log_side_slope = params.log_side_slope
Expand All @@ -218,29 +225,16 @@ def apply_log_internal(value: NDArrayFloat, params, extra_args) -> NDArrayFloat:
lin_side_break = params.lin_side_break
base = params.base
linear_slope = params.linear_slope
if lin_side_slope is None:
linear_slope = (
log_side_slope
* lin_side_slope
/ (
(lin_side_slope * lin_side_break + lin_side_offset)
* np.log(base)
)
)
log_side_break = (
log_side_slope
* emath.logn(base, lin_side_slope * lin_side_break + lin_side_offset)
+ log_side_offset
)
linear_offset = log_side_break - linear_slope * lin_side_break
return np.where(
value <= log_side_break,
(value - linear_offset) / linear_slope,
(
(base ** ((value - log_side_offset) / log_side_slope))
- lin_side_offset
)
/ lin_side_slope,
return logarithmic_function_camera(
value,
"cameraLogToLin",
base,
log_side_slope,
lin_side_slope,
log_side_offset,
lin_side_offset,
lin_side_break,
linear_slope,
)
case _:
raise ValueError(f"Invalid Log Style: {style}")
Expand Down Expand Up @@ -282,44 +276,32 @@ def apply_exponent_internal(
style = extra_args
match style:
case ExponentStyle.BASIC_FWD:
return np.maximum(0.0, value) ** exponent
return exponent_function_basic(value, exponent, "basicFwd")
case ExponentStyle.BASIC_REV:
return np.maximum(0.0, value) ** (1 / exponent)
return exponent_function_basic(value, exponent, "basicRev")
case ExponentStyle.BASIC_MIRROR_FWD:
return np.where(
value >= 0,
value**exponent,
-((-value) ** exponent), # TODO check for type in reference
)
return exponent_function_basic(value, exponent, "basicMirrorFwd")
case ExponentStyle.BASIC_MIRROR_REV:
return np.where(
value >= 0,
value ** (1 / exponent),
-((-value) ** (1 / exponent)),
)
return exponent_function_basic(value, exponent, "basicMirrorRev")
case ExponentStyle.BASIC_PASS_THRU_FWD:
return np.where(value >= 0, value**exponent, value)
return exponent_function_basic(value, exponent, "basicPassThruFwd")
case ExponentStyle.BASIC_PASS_THRU_REV:
return np.where(
value >= 0,
value ** (1 / exponent),
value,
)
return exponent_function_basic(value, exponent, "basicPassThruRev")
case ExponentStyle.MON_CURVE_FWD:
return mon_curve_forward(value, exponent, offset)
return exponent_function_monitor_curve(
value, exponent, offset, "monCurveFwd"
)
case ExponentStyle.MON_CURVE_REV:
return mon_curve_reverse(value, exponent, offset)
return exponent_function_monitor_curve(
value, exponent, offset, "monCurveRev"
)
case ExponentStyle.MON_CURVE_MIRROR_FWD:
return np.where(
value >= 0,
mon_curve_forward(value, exponent, offset),
-mon_curve_forward(-value, exponent, offset),
return exponent_function_monitor_curve(
value, exponent, offset, "monCurveMirrorFwd"
)
case ExponentStyle.MON_CURVE_MIRROR_REV:
return np.where(
value >= 0,
mon_curve_reverse(value, exponent, offset),
-mon_curve_reverse(-value, exponent, offset),
return exponent_function_monitor_curve(
value, exponent, offset, "monCurveMirrorRev"
)
case _:
raise ValueError(f"Invalid Exponent Style: {style}")
Expand All @@ -334,8 +316,10 @@ def apply_exponent(node: clf.Exponent, normalised_value: NDArrayFloat) -> NDArra


def asc_cdl_luma(value):
R, G, B = tsplit(value)
luma = 0.2126 * R + 0.7152 * G + 0.0722 * B
# R, G, B = tsplit(value)
# luma = 0.2126 * R + 0.7152 * G + 0.0722 * B
weights = [0.2126, 0.7152, 0.0722]
luma = np.sum(weights * value, axis=-1)
return luma


Expand Down

0 comments on commit 77c42f8

Please sign in to comment.