[Bf-blender-cvs] SVN commit: /data/svn/bf-blender [49134] branches/soc-2011-tomato/source/ blender/imbuf/intern/filter.c: ACES ODT tonecurve style cleanup and small tweaks
Sergey Sharybin
sergey.vfx at gmail.com
Mon Jul 23 11:20:11 CEST 2012
Revision: 49134
http://projects.blender.org/scm/viewvc.php?view=rev&root=bf-blender&revision=49134
Author: nazgul
Date: 2012-07-23 09:20:11 +0000 (Mon, 23 Jul 2012)
Log Message:
-----------
ACES ODT tonecurve style cleanup and small tweaks
Convert color into ACES color space from rec709 (which is quite the same
as blender internal) before running tonecurve and convert color space back
after tonecurve.
This makes tonecurve be applying in actual space it was intended to and it
makes color range smoother a bit.
Would be nice to convert this into OCIO view tho.
Modified Paths:
--------------
branches/soc-2011-tomato/source/blender/imbuf/intern/filter.c
Modified: branches/soc-2011-tomato/source/blender/imbuf/intern/filter.c
===================================================================
--- branches/soc-2011-tomato/source/blender/imbuf/intern/filter.c 2012-07-23 09:18:32 UTC (rev 49133)
+++ branches/soc-2011-tomato/source/blender/imbuf/intern/filter.c 2012-07-23 09:20:11 UTC (rev 49134)
@@ -31,7 +31,6 @@
* \ingroup imbuf
*/
-#define _GNU_SOURCE /* See feature_test_macros(7) */
#include <math.h>
#include "MEM_guardedalloc.h"
@@ -601,185 +600,215 @@
/* Tonecurve corrections */
-// code of rdt_shaper_fwd and ratio_preserving_odt_tonecurve belongs to
-// ACES project (https://github.com/ampas/aces-dev)
+/* code of rdt_shaper_fwd and ratio_preserving_odt_tonecurve belongs to
+ * ACES project (https://github.com/ampas/aces-dev)
+ */
-// === ODT SPLINE === //
-//
-// Algorithm for applying ODT tone curve in forward direction.
-//
-// vers 1.0 Doug Walker 2012-01-23
-// modified by Scott Dyer 2012-02-28
+/* === ODT SPLINE === */
+/*
+ * Algorithm for applying ODT tone curve in forward direction.
+ *
+ * vers 1.0 Doug Walker 2012-01-23
+ * modified by Scott Dyer 2012-02-28
+ */
-// Input and output are in linear (not log) units.
-static float rdt_shaper_fwd( float x)
+/* Input and output are in linear (not log) units. */
+static float rdt_shaper_fwd(float x)
{
- // B-spline coefficients.
- // The units are density of the output.
- const float COEFS0 = -0.008;
- const float COEFS1 = -0.00616;
- const float COEFS2 = 0.026;
- const float COEFS3 = 0.185;
- const float COEFS4 = 0.521;
- const float COEFS5 = 0.993;
- const float COEFS6 = 1.563;
- const float COEFS7 = 2.218;
- const float COEFS8 = 2.795;
- const float COEFS9 = 3.36;
- const float COEFS10 = 4.0; // NB: keep this less than or equal to -log10( FLARE)
- // The locations of these control points in OCES density space are:
- // -1., -0.79, -0.44, -0.01, 0.48, 1.01, 1.58, 2.18, 2.82, 3.47, 4.15, 4.85
+ /* B-spline coefficients.
+ * The units are density of the output.
+ */
+ const float COEFS0 = -0.008f;
+ const float COEFS1 = -0.00616f;
+ const float COEFS2 = 0.026f;
+ const float COEFS3 = 0.185f;
+ const float COEFS4 = 0.521f;
+ const float COEFS5 = 0.993f;
+ const float COEFS6 = 1.563f;
+ const float COEFS7 = 2.218f;
+ const float COEFS8 = 2.795f;
+ const float COEFS9 = 3.36f;
+ const float COEFS10 = 4.0f; /* NB: keep this less than or equal to -log10(FLARE) */
+ /* The locations of these control points in OCES density space are:
+ * -1., -0.79, -0.44, -0.01, 0.48, 1.01, 1.58, 2.18, 2.82, 3.47, 4.15, 4.85
+ */
- // The flare term allows the spline to more rapidly approach zero
- // while keeping the shape of the curve well-behaved in density space.
- const float FLARE = 1e-4;
+ /* The flare term allows the spline to more rapidly approach zero
+ * while keeping the shape of the curve well-behaved in density space.
+ */
+ const float FLARE = 1e-4f;
- // The last control point is fixed to yield a specific density at the
- // end of the knot domain.
- //const float COEFS11 = 2. * ( -log10( FLARE) - 0.001) - COEFS10;
- // Note: Apparently a CTL bug prevents calling log10() here, so
- // you'll need to update this manually if you change FLARE.
- const float COEFS11 = COEFS10 + 2. * ( 4. - COEFS10);
+ /* The last control point is fixed to yield a specific density at the
+ * end of the knot domain.
+ */
+ /* const float COEFS11 = 2. * (-log10(FLARE) - 0.001) - COEFS10; */
- // The knots are in units of OCES density.
+ /* Note: Apparently a CTL bug prevents calling log10() here, so
+ * you'll need to update this manually if you change FLARE.
+ */
+ const float COEFS11 = COEFS10 + 2.0f * (4.0f - COEFS10);
+
+ /* The knots are in units of OCES density. */
const unsigned int KNOT_LEN = 11;
- const float KNOT_START = -0.9;
- const float KNOT_END = 4.484256;
+ const float KNOT_START = -0.9f;
+ const float KNOT_END = 4.484256f;
- // The KNOT_POW adjusts the spacing to put more knots near the toe (highlights).
- const float KNOT_POW = 1. / 1.3;
+ /* The KNOT_POW adjusts the spacing to put more knots near the toe (highlights). */
+ const float KNOT_POW = 1.0f / 1.3f;
const float OFFS = KNOT_START;
const float SC = KNOT_END - KNOT_START;
- // KNOT_DENS is density of the spline at the knots.
- const float KNOT_DENS[ 11] = {
- ( COEFS0 + COEFS1) / 2.,
- ( COEFS1 + COEFS2) / 2.,
- ( COEFS2 + COEFS3) / 2.,
- ( COEFS3 + COEFS4) / 2.,
- ( COEFS4 + COEFS5) / 2.,
- ( COEFS5 + COEFS6) / 2.,
- ( COEFS6 + COEFS7) / 2.,
- ( COEFS7 + COEFS8) / 2.,
- ( COEFS8 + COEFS9) / 2.,
- ( COEFS9 + COEFS10) / 2.,
- ( COEFS10 + COEFS11) / 2.
+ /* KNOT_DENS is density of the spline at the knots. */
+ const float KNOT_DENS[11] = {
+ (COEFS0 + COEFS1) / 2.0f,
+ (COEFS1 + COEFS2) / 2.0f,
+ (COEFS2 + COEFS3) / 2.0f,
+ (COEFS3 + COEFS4) / 2.0f,
+ (COEFS4 + COEFS5) / 2.0f,
+ (COEFS5 + COEFS6) / 2.0f,
+ (COEFS6 + COEFS7) / 2.0f,
+ (COEFS7 + COEFS8) / 2.0f,
+ (COEFS8 + COEFS9) / 2.0f,
+ (COEFS9 + COEFS10) / 2.0f,
+ (COEFS10 + COEFS11) / 2.0f
};
- // Parameters controlling linear extrapolation.
+ /* Parameters controlling linear extrapolation. */
const float LIGHT_SLOPE = 0.023;
const float CROSSOVER = pow(10,-KNOT_END);
const float REV_CROSSOVER = pow(10.0f, -KNOT_DENS[ KNOT_LEN - 1]) - FLARE;
const float DARK_SLOPE = REV_CROSSOVER / CROSSOVER;
- // Textbook monomial to basis-function conversion matrix.
- /*const*/ float M[ 3][ 3] = {
- { 0.5, -1.0, 0.5 },
- { -1.0, 1.0, 0.5 },
- { 0.5, 0.0, 0.0 }
+ /* Textbook monomial to basis-function conversion matrix. */
+ /*const*/ float M[3][3] = {
+ { 0.5f, -1.0f, 0.5f},
+ {-1.0f, 1.0f, 0.5f},
+ { 0.5f, 0.0f, 0.0f}
};
- float y;
- // Linear extrapolation in linear space for negative & very dark values.
- if ( x <= CROSSOVER)
- y = x * DARK_SLOPE;
- else {
- float in_dens = -log10( x);
- float out_dens;
- float knot_coord = ( in_dens - OFFS) / SC;
+ float y;
- // Linear extrapolation in log space for very light values.
- if ( knot_coord <= 0.)
- out_dens = KNOT_DENS[ 0] - ( KNOT_START - in_dens) * LIGHT_SLOPE;
+ /* Linear extrapolation in linear space for negative & very dark values. */
+ if (x <= CROSSOVER) {
+ y = x * DARK_SLOPE;
+ }
+ else {
+ float in_dens = -log10(x);
+ float out_dens;
+ float knot_coord = (in_dens - OFFS) / SC;
- // For typical OCES values, apply a B-spline curve.
- else {
- knot_coord = ( KNOT_LEN - 1) * pow( knot_coord, KNOT_POW);
- {
- int j = knot_coord;
- float t = knot_coord - j;
+ if (knot_coord <= 0.0f) {
+ /* Linear extrapolation in log space for very light values. */
- // Would like to do this:
- //float cf[ 3] = { COEFS[ j], COEFS[ j + 1], COEFS[ j + 2]};
- // or at least:
- //cf[ 0] = COEFS[ j];
- //cf[ 1] = COEFS[ j + 1];
- //cf[ 2] = COEFS[ j + 2];
- // But apparently CTL bugs prevent it, so we do the following:
- float cf[ 3];
- if ( j <= 0) {
- cf[ 0] = COEFS0; cf[ 1] = COEFS1; cf[ 2] = COEFS2;
- }
- else if ( j == 1) {
- cf[ 0] = COEFS1; cf[ 1] = COEFS2; cf[ 2] = COEFS3;
- }
- else if ( j == 2) {
- cf[ 0] = COEFS2; cf[ 1] = COEFS3; cf[ 2] = COEFS4;
- }
- else if ( j == 3) {
- cf[ 0] = COEFS3; cf[ 1] = COEFS4; cf[ 2] = COEFS5;
- }
- else if ( j == 4) {
- cf[ 0] = COEFS4; cf[ 1] = COEFS5; cf[ 2] = COEFS6;
- }
- else if ( j == 5) {
- cf[ 0] = COEFS5; cf[ 1] = COEFS6; cf[ 2] = COEFS7;
- }
- else if ( j == 6) {
- cf[ 0] = COEFS6; cf[ 1] = COEFS7; cf[ 2] = COEFS8;
- }
- else if ( j == 7) {
- cf[ 0] = COEFS7; cf[ 1] = COEFS8; cf[ 2] = COEFS9;
- }
- else if ( j == 8) {
- cf[ 0] = COEFS8; cf[ 1] = COEFS9; cf[ 2] = COEFS10;
- }
- else {
- cf[ 0] = COEFS9; cf[ 1] = COEFS10; cf[ 2] = COEFS11;
- }
+ out_dens = KNOT_DENS[0] - (KNOT_START - in_dens) * LIGHT_SLOPE;
+ }
+ else {
+ /* For typical OCES values, apply a B-spline curve. */
- {
- float monomials[ 3] = { t * t, t, 1. };
- float v[3];
+ int j;
+ float t;
+ float cf[3], monomials[3], v[3];
- // XXX: check on this! maths could be different here (like row-major vs. column major or so)
- //out_dens = dot_f3_f3( monomials, mult_f3_f33( cf, M));
+ knot_coord = ( KNOT_LEN - 1) * pow( knot_coord, KNOT_POW);
- mul_v3_m3v3(v, M, cf);
- out_dens = dot_v3v3( monomials, v);
- }
+ j = knot_coord;
+ t = knot_coord - j;
+
+ /* Would like to do this: */
+ /* float cf[ 3] = { COEFS[ j], COEFS[ j + 1], COEFS[ j + 2]}; */
+ /* or at least: */
+ /* cf[0] = COEFS[j]; */
+ /* cf[1] = COEFS[j + 1]; */
+ /* cf[2] = COEFS[j + 2]; */
+ /* But apparently CTL bugs prevent it, so we do the following: */
+ if (j <= 0) {
+ cf[0] = COEFS0; cf[1] = COEFS1; cf[2] = COEFS2;
}
- }
- y = pow(10.0f, -out_dens) - FLARE;
- }
- return y;
+ else if (j == 1) {
+ cf[0] = COEFS1; cf[1] = COEFS2; cf[2] = COEFS3;
+ }
+ else if (j == 2) {
+ cf[0] = COEFS2; cf[1] = COEFS3; cf[2] = COEFS4;
+ }
+ else if (j == 3) {
+ cf[0] = COEFS3; cf[1] = COEFS4; cf[2] = COEFS5;
+ }
+ else if (j == 4) {
+ cf[0] = COEFS4; cf[1] = COEFS5; cf[2] = COEFS6;
+ }
+ else if (j == 5) {
+ cf[0] = COEFS5; cf[1] = COEFS6; cf[2] = COEFS7;
+ }
+ else if (j == 6) {
+ cf[0] = COEFS6; cf[1] = COEFS7; cf[2] = COEFS8;
+ }
+ else if (j == 7) {
+ cf[0] = COEFS7; cf[1] = COEFS8; cf[2] = COEFS9;
+ }
+ else if (j == 8) {
+ cf[0] = COEFS8; cf[1] = COEFS9; cf[2] = COEFS10;
+ }
+ else {
+ cf[0] = COEFS9; cf[1] = COEFS10; cf[2] = COEFS11;
+ }
+
+ monomials[0] = t * t;
+ monomials[1] = t;
+ monomials[2] = 1.0f;
+
+ mul_v3_m3v3(v, M, cf);
+ out_dens = dot_v3v3( monomials, v);
+ }
+
+ y = pow(10.0f, -out_dens) - FLARE;
+ }
+
+ return y;
}
void IMB_ratio_preserving_odt_tonecurve(float rgbOut[3], const float rgbIn[3])
{
- //
- // The "ratio preserving tonecurve" is used to avoid hue/chroma shifts.
- // It sends a norm through the tonecurve and scales the RGB values based on the output.
- //
+ float acesIn[3], acesOut[3];
- const float NTH_POWER = 2.0;
- const float TINY = 1e-12;
+ float rec709_to_aces[3][3] = {{0.4395770431f, 0.0895979404f, 0.0174174830f},
+ {0.3839148879f, 0.8147120476f, 0.1087378338f},
@@ Diff output truncated at 10240 characters. @@
More information about the Bf-blender-cvs
mailing list