jp9000/obs-studio
1
0
Fork 0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

obs-filters: Interpolate LUT in linear space

Browse Source 
More accurate and cheaper.

This is a breaking change because Amount is also now computed as linear,
but most users are probably using 1.0, so v2 isn't worthwhile.

Intel UHD Graphics 750: 980 µs -> 860 µs
This commit is contained in:
jpark37 2022-02-16 23:12:26 -08:00 committed by Jim
parent c434d4d39b
commit 8e8c29d84c
2 changed files with 27 additions and 32 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

22
plugins/obs-filters/color-grade-filter.c
View File

@ -126,9 +126,12 @@ static void *load_1d_lut(FILE *const file, const uint32_t width, float red,
break;
}
values[offset++] = half_from_float(red);
values[offset++] = half_from_float(green);
values[offset++] = half_from_float(blue);
values[offset++] =
half_from_float(gs_srgb_nonlinear_to_linear(red));
values[offset++] =
half_from_float(gs_srgb_nonlinear_to_linear(green));
values[offset++] =
half_from_float(gs_srgb_nonlinear_to_linear(blue));
values[offset++] = half_from_bits(0x3c00); // 1.0
data_found = get_cube_entry(file, &red, &green, &blue);
@ -155,9 +158,12 @@ static void *load_3d_lut(FILE *const file, const uint32_t width, float red,
break;
}
values[offset++] = half_from_float(red);
values[offset++] = half_from_float(green);
values[offset++] = half_from_float(blue);
values[offset++] = half_from_float(
gs_srgb_nonlinear_to_linear(red));
values[offset++] = half_from_float(
gs_srgb_nonlinear_to_linear(green));
values[offset++] = half_from_float(
gs_srgb_nonlinear_to_linear(blue));
values[offset++] =
half_from_bits(0x3c00); // 1.0
@ -374,7 +380,7 @@ static obs_properties_t *color_grade_filter_properties(void *data)
obs_properties_add_path(props, SETTING_IMAGE_PATH, TEXT_IMAGE_PATH,
OBS_PATH_FILE, filter_str.array, path.array);
obs_properties_add_float_slider(props, SETTING_CLUT_AMOUNT, TEXT_AMOUNT,
0, 1, 0.01);
0, 1, 0.0001);
dstr_free(&filter_str);
dstr_free(&path);
@ -432,7 +438,7 @@ static void color_grade_filter_render(void *data, gs_effect_t *effect)
}
param = gs_effect_get_param_by_name(filter->effect, clut_texture_name);
gs_effect_set_texture(param, filter->target);
gs_effect_set_texture_srgb(param, filter->target);
param = gs_effect_get_param_by_name(filter->effect, "clut_amount");
gs_effect_set_float(param, filter->clut_amount);

37
plugins/obs-filters/data/color_grade_filter.effect
View File

@ -45,41 +45,30 @@ float3 srgb_linear_to_nonlinear(float3 v)
return float3(srgb_linear_to_nonlinear_channel(v.r), srgb_linear_to_nonlinear_channel(v.g), srgb_linear_to_nonlinear_channel(v.b));
}
float srgb_nonlinear_to_linear_channel(float u)
{
return (u <= 0.04045) ? (u / 12.92) : pow((u + 0.055) / 1.055, 2.4);
}
float3 srgb_nonlinear_to_linear(float3 v)
{
return float3(srgb_nonlinear_to_linear_channel(v.r), srgb_nonlinear_to_linear_channel(v.g), srgb_nonlinear_to_linear_channel(v.b));
}
float4 LUT1D(VertDataOut v_in) : TARGET
{
float4 textureColor = image.Sample(textureSampler, v_in.uv);
textureColor.rgb = max(float3(0.0, 0.0, 0.0), textureColor.rgb / textureColor.a);
textureColor.rgb = srgb_linear_to_nonlinear(textureColor.rgb);
float3 nonlinear = srgb_linear_to_nonlinear(textureColor.rgb);
if (textureColor.r >= domain_min.r && textureColor.r <= domain_max.r) {
float u = textureColor.r * clut_scale.r + clut_offset.r;
if (nonlinear.r >= domain_min.r && nonlinear.r <= domain_max.r) {
float u = nonlinear.r * clut_scale.r + clut_offset.r;
float channel = clut_1d.Sample(textureSampler, float2(u, 0.5)).r;
textureColor.r = lerp(textureColor.r, channel, clut_amount);
}
if (textureColor.g >= domain_min.g && textureColor.g <= domain_max.g) {
float u = textureColor.g * clut_scale.g + clut_offset.g;
if (nonlinear.g >= domain_min.g && nonlinear.g <= domain_max.g) {
float u = nonlinear.g * clut_scale.g + clut_offset.g;
float channel = clut_1d.Sample(textureSampler, float2(u, 0.5)).g;
textureColor.g = lerp(textureColor.g, channel, clut_amount);
}
if (textureColor.b >= domain_min.b && textureColor.b <= domain_max.b) {
float u = textureColor.b * clut_scale.b + clut_offset.b;
if (nonlinear.b >= domain_min.b && nonlinear.b <= domain_max.b) {
float u = nonlinear.b * clut_scale.b + clut_offset.b;
float channel = clut_1d.Sample(textureSampler, float2(u, 0.5)).b;
textureColor.b = lerp(textureColor.b, channel, clut_amount);
}
textureColor.rgb = srgb_nonlinear_to_linear(textureColor.rgb);
return textureColor;
}
@ -87,15 +76,16 @@ float4 LUT3D(VertDataOut v_in) : TARGET
{
float4 textureColor = image.Sample(textureSampler, v_in.uv);
textureColor.rgb = max(float3(0.0, 0.0, 0.0), textureColor.rgb / textureColor.a);
textureColor.rgb = srgb_linear_to_nonlinear(textureColor.rgb);
float r = textureColor.r;
float g = textureColor.g;
float b = textureColor.b;
float3 nonlinear = srgb_linear_to_nonlinear(textureColor.rgb);
float r = nonlinear.r;
float g = nonlinear.g;
float b = nonlinear.b;
if (r >= domain_min.r && r <= domain_max.r &&
g >= domain_min.g && g <= domain_max.g &&
b >= domain_min.b && b <= domain_max.b)
{
float3 clut_pos = textureColor.rgb * clut_scale + clut_offset;
float3 clut_pos = nonlinear * clut_scale + clut_offset;
float3 floor_pos = floor(clut_pos);
float3 fracRGB = clut_pos - floor_pos;
@ -170,7 +160,6 @@ float4 LUT3D(VertDataOut v_in) : TARGET
textureColor.rgb = lerp(textureColor.rgb, luttedColor, clut_amount);
}
textureColor.rgb = srgb_nonlinear_to_linear(textureColor.rgb);
textureColor.rgb *= textureColor.a;
return textureColor;
}