OpenColorIO

Open Source Color Management

OpenColorIO v1.0.8 documentation

Usage Examples

Some examples of using the OCIO API, via both C++ and the Python bindings.

For further examples, see the src/apps/ directory in the git repository

Applying a basic ColorSpace transform, using the CPU

This describes what code is used to convert from a specified source ColorSpace to a specified destination ColorSpace. If you are using the OCIO Nuke plugins, the OCIOColorSpace node performs these steps internally.

  1. Get the Config. This represents the entirety of the current color “universe”. It can either be initialized by your app at startup or created explicitly. In common usage, you can just query GetCurrentConfig(), which will auto initialize on first use using the OCIO environment variable.
  2. Get Processor from the Config. A processor corresponds to a ‘baked’ color transformation. You specify two arguments when querying a processor: the ColorSpace you are coming from, and the ColorSpace you are going to. ColorSpaces ColorSpaces can be either explicitly named strings (defined by the current configuration) or can be Roles (essentially ColorSpace aliases) which are consistent across configurations. Constructing a Processor object is likely a blocking operation (thread-wise) so care should be taken to do this as infrequently as is sensible. Once per render ‘setup’ would be appropriate, once per scanline would be inappropriate.
  3. Convert your image, using the Processor. Once you have the processor, you can apply the color transformation using the “apply” function. In C++, you apply the processing in-place, by first wrapping your image in an ImageDesc class. This approach is intended to be used in high performance applications, and can be used on multiple threads (per scanline, per tile, etc). In Python you call “applyRGB” / “applyRGBA” on your sequence of pixels. Note that in both languages, it is far more efficient to call “apply” on batches of pixels at a time.

C++

#include <OpenColorIO/OpenColorIO.h>
namespace OCIO = OCIO_NAMESPACE;

try
{
    OCIO::ConstConfigRcPtr config = OCIO::GetCurrentConfig();
    ConstProcessorRcPtr processor = config->getProcessor(OCIO::ROLE_COMPOSITING_LOG,
                                                         OCIO::ROLE_SCENE_LINEAR);

    OCIO::PackedImageDesc img(imageData, w, h, 4);
    processor->apply(img);
}
catch(OCIO::Exception & exception)
{
    std::cerr << "OpenColorIO Error: " << exception.what() << std::endl;
}

Python

import PyOpenColorIO as OCIO

try:
    config = OCIO.GetCurrentConfig()
    processor = config.getProcessor(OCIO.Constants.ROLE_COMPOSITING_LOG,
                                    OCIO.Constants.ROLE_SCENE_LINEAR)

    # Apply the color transform to the existing RGBA pixel data
    img = processor.applyRGBA(img)
except Exception, e:
    print "OpenColorIO Error",e

Displaying an image, using the CPU (simple ColorSpace conversion)

Converting an image for display is similar to a normal color space conversion. The only difference is that one has to first determine the name of the display (destination) ColorSpace by querying the config with the device name and transform name.

  1. Get the Config. See Applying a basic ColorSpace transform, using the CPU for details.
  2. Lookup the display ColorSpace. The display ColorSpace is queried from the configuration using Config::getDisplayColorSpaceName(). If the user has specified value for the device or the displayTransformName, use them. If these values are unknown default values can be queried (as shown below).
  3. Get the processor from the Config. See Applying a basic ColorSpace transform, using the CPU for details.
  4. Convert your image, using the processor. See Applying a basic ColorSpace transform, using the CPU for details.

C++

#include <OpenColorIO/OpenColorIO.h>
namespace OCIO = OCIO_NAMESPACE;

OCIO::ConstConfigRcPtr config = OCIO::GetCurrentConfig();

// If the user hasn't picked a display, use the defaults...
const char * device = config->getDefaultDisplayDeviceName();
const char * transformName = config->getDefaultDisplayTransformName(device);
const char * displayColorSpace = config->getDisplayColorSpaceName(device, transformName);

ConstProcessorRcPtr processor = config->getProcessor(OCIO::ROLE_SCENE_LINEAR,
                                                     displayColorSpace);

OCIO::PackedImageDesc img(imageData, w, h, 4);
processor->apply(img);

Python

import PyOpenColorIO as OCIO

config = OCIO.GetCurrentConfig()

device = config.getDefaultDisplayDeviceName()
transformName = config.getDefaultDisplayTransformName(device)
displayColorSpace = config.getDisplayColorSpaceName(device, transformName)

processor = config.getProcessor(OCIO.Constants.ROLE_SCENE_LINEAR, displayColorSpace)

processor.applyRGB(imageData)

Displaying an image, using the CPU (Full Display Pipeline)

This alternative version allows for a more complex displayTransform, allowing for all of the controls typically added to real-world viewer interfaces. For example, options are allowed to control which channels (red, green, blue, alpha, luma) are visible, as well as allowing for optional color corrections (such as an exposure offset in scene linear). If you are using the OCIO Nuke plugins, the OCIODisplay node performs these steps internally.

  1. Get the Config. See Applying a basic ColorSpace transform, using the CPU for details.
  2. Lookup the display ColorSpace. See Displaying an image, using the CPU (simple ColorSpace conversion) for details
  3. Create a new DisplayTransform. This transform will embody the full ‘display’ pipeline you wish to control. The user is required to call DisplayTransform::setInputColorSpaceName() to set the input ColorSpace, as well as DisplayTransform::setDisplayColorSpaceName() (with the results of Config::getDisplayColorSpaceName()).
  4. Set any additional DisplayTransform options. If the user wants to specify a channel swizzle, a scene-linear exposure offset, an artistic look, this is the place to add it. See below for an example. Note that although we provide recommendations for display, any transforms are allowed to be added into any of the slots. So if for your app you want to add 3 transforms into a particular slot (chained together), you are free to wrap them in a GroupTransform and set it accordingly!
  5. Get the processor from the Config. The processor is then queried from the config passing the new DisplayTransform as the argument. Once the processor has been returned, the original DisplayTransform is no longer necessary to hold onto. (Though if you’d like to for re-use, there is no problem doing so).
  6. Convert your image, using the processor. See Applying a basic ColorSpace transform, using the CPU for details.

C++

// Step 1: Get the config
OCIO::ConstConfigRcPtr config = OCIO::GetCurrentConfig();

// Step 2: Lookup the display ColorSpace
const char * device = config->getDefaultDisplayDeviceName();
const char * transformName = config->getDefaultDisplayTransformName(device);
const char * displayColorSpace = config->getDisplayColorSpaceName(device, transformName);

// Step 3: Create a DisplayTransform, and set the input and display ColorSpaces
// (This example assumes the input is scene linear. Adapt as needed.)

OCIO::DisplayTransformRcPtr transform = OCIO::DisplayTransform::Create();
transform->setInputColorSpaceName( OCIO::ROLE_SCENE_LINEAR );
transform->setDisplayColorSpaceName( displayColorSpace );

// Step 4: Add custom transforms for a 'canonical' Display Pipeline

// Add an fstop exposure control (in SCENE_LINEAR)
float gain = powf(2.0f, exposure_in_stops);
const float slope3f[] = { gain, gain, gain };
OCIO::CDLTransformRcPtr cc =  OCIO::CDLTransform::Create();
cc->setSlope(slope3f);
transform->setLinearCC(cc);

// Add a Channel view 'swizzle'

// 'channelHot' controls which channels are viewed.
int channelHot[4] = { 1, 1, 1, 1 };  // show rgb
//int channelHot[4] = { 1, 0, 0, 0 };  // show red
//int channelHot[4] = { 0, 0, 0, 1 };  // show alpha
//int channelHot[4] = { 1, 1, 1, 0 };  // show luma

float lumacoef[3];
config.getDefaultLumaCoefs(lumacoef);

float m44[16];
float offset[4];
OCIO::MatrixTransform::View(m44, offset, channelHot, lumacoef);
OCIO::MatrixTransformRcPtr swizzle = OCIO::MatrixTransform::Create();
swizzle->setValue(m44, offset);
transform->setChannelView(swizzle);

// And then process the image normally.
OCIO::ConstProcessorRcPtr processor = config->getProcessor(transform);

OCIO::PackedImageDesc img(imageData, w, h, 4);
processor->apply(img);

Python

import PyOpenColorIO as OCIO

# Step 1: Get the config
config = OCIO.GetCurrentConfig()

# Step 2: Lookup the display ColorSpace
device = config.getDefaultDisplayDeviceName()
transformName = config.getDefaultDisplayTransformName(device)
displayColorSpace = config.getDisplayColorSpaceName(device, transformName)

# Step 3: Create a DisplayTransform, and set the input and display ColorSpaces
# (This example assumes the input is scene linear. Adapt as needed.)

transform = OCIO.DisplayTransform()
transform.setInputColorSpaceName(OCIO.Constants.ROLE_SCENE_LINEAR)
transform.setDisplayColorSpaceName(displayColorSpace)

# Step 4: Add custom transforms for a 'canonical' Display Pipeline

# Add an fstop exposure control (in SCENE_LINEAR)
gain = 2**exposure
slope3f = (gain, gain, gain)

cc = OCIO.CDLTransform()
cc.setSlope(slope3f)

transform.setLinearCC(cc)

# Add a Channel view 'swizzle'

channelHot = (1, 1, 1, 1) # show rgb
# channelHot = (1, 0, 0, 0) # show red
# channelHot = (0, 0, 0, 1) # show alpha
# channelHot = (1, 1, 1, 0) # show luma

lumacoef = config.getDefaultLumaCoefs()

m44, offset = OCIO.MatrixTransform.View(channelHot, lumacoef)

swizzle = OCIO.MatrixTransform()
swizzle.setValue(m44, offset)
transform.setChannelView(swizzle)

# And then process the image normally.
processor = config.getProcessor(transform)

print processor.applyRGB(imageData)

Displaying an image, using the GPU

Applying OpenColorIO’s color processing using GPU processing is straightforward, provided you have the capability to upload custom shader code and a custom 3D Lookup Table (3DLUT).

  1. Get the Processor. This portion of the pipeline is identical to the CPU approach. Just get the processor as you normally would have, see above for details.
  2. Create a GpuShaderDesc.
  3. Query the GPU Shader Text + 3D LUT.
  4. Configure the GPU State.
  5. Draw your image.

C++

This example is available as a working app in the OCIO source: src/apps/ociodisplay.

// Step 0: Get the processor using any of the pipelines mentioned above.
OCIO::ConstConfigRcPtr config = OCIO::GetCurrentConfig();
const char * device = config->getDefaultDisplayDeviceName();
const char * transformName = config->getDefaultDisplayTransformName(device);
const char * displayColorSpace = config->getDisplayColorSpaceName(device, transformName);
ConstProcessorRcPtr processor = config->getProcessor(OCIO::ROLE_SCENE_LINEAR,
                                                     displayColorSpace);

// Step 1: Create a GPU Shader Description
GpuShaderDesc shaderDesc;
shaderDesc.setLanguage(OCIO::GPU_LANGUAGE_GLSL_1_0);
shaderDesc.setFunctionName("OCIODisplay");
const int LUT3D_EDGE_SIZE = 32;
shaderDesc.setLut3DEdgeLen(LUT3D_EDGE_SIZE);

// Step 2: Compute and the 3D LUT
// Optional Optimization:
//     Only do this the 3D LUT's contents
//     are different from the last drawn frame.
//     Use getGpuLut3DCacheID to compute the cacheID.
//     cheaply.
//
// std::string lut3dCacheID = processor->getGpuLut3DCacheID(shaderDesc);
int num3Dentries = 3*LUT3D_EDGE_SIZE*LUT3D_EDGE_SIZE*LUT3D_EDGE_SIZE;
std::vector<float> g_lut3d;
g_lut3d.resize(num3Dentries);
processor->getGpuLut3D(&g_lut3d[0], shaderDesc);

// Load the data into an OpenGL 3D Texture
glGenTextures(1, &g_lut3d_textureID);
glBindTexture(GL_TEXTURE_3D, g_lut3d_textureID);
glTexImage3D(GL_TEXTURE_3D, 0, GL_RGB,
             LUT3D_EDGE_SIZE, LUT3D_EDGE_SIZE, LUT3D_EDGE_SIZE,
             0, GL_RGB,GL_FLOAT, &g_lut3d[0]);

// Step 3: Query