|const IceTVoid *||color_buffer,|
|const IceTVoid *||depth_buffer,|
|const IceTInt *||valid_pixels_viewport,|
|const IceTDouble *||projection_matrix,|
|const IceTDouble *||modelview_matrix,|
|const IceTFloat *||background_color );|
The icetCompositeImage function takes image buffer data and composites it to a single image. This function behaves similarly to icetDrawFrame except that instead of using callback functions to render and retrieve image data, the images are pre-rendered and passed directly to icetCompositeImage. Although it is more efficient to allow IceT to determine rendering projections and use callbacks, it is often more convenient for applications to integrate IceT as a separate compositing step after the rendering.
Before IceT may composite an image, the display needs to be defined (using icetAddTile), the buffer formats need to be specified (using icetSetColorFormatand icetSetDepthFormat),and the composite strategy must be set (using icetStrategy). The single image sub-strategy may also optionally be set (using icetSingleImageStrategy).
All process must call icetCompositeImage for the operation to complete on any process in a parallel job.
The color_buffer and depth_buffer arguments point to memory buffers that contain the image data. The image data is always stored in densely packed arrays in row-major order (a.k.a. x-major order or by scan-lines). The first horizontal scan-line is at the bottom of the image with subsequent scan-lines moving up. The size of each image buffer is expected to be the width and the height of the global viewport (which is set indirectly with icetAddTile). The global viewport is stored in the ICET_GLOBAL_VIEWPORT state variable. If only one tile is specified, then the width and height of the global viewport will be the same as this one tile.
The format for color_buffer is expected to be the same as what is set with icetSetColorFormat.The following formats and their interpretations with respect to color_buffer are as follows.
Each entry is an RGBA color tuple. Each component is valued in the range from 0 to 255 and is stored as an 8-bit integer. The buffer will always be allocated on memory boundaries such that each color value can be treated as a single 32-bit integer.
Each entry is an RGBA color tuple. Each component is in the range from 0.0 to 1.0 and is stored as a 32-bit float.
No color values are stored in the image.
Likewise, the format for depth_buffer is expected to be the same as what is set with icetSetDepthFormat.The following formats and their interpretations with respect to depth_buffer are as follows.
Each entry is in the range from 0.0 (near plane) to 1.0 (far plane) and is stored as a 32-bit float.
No depth values are stored in the image.
If the current format does not have a color or depth, then the respective buffer argument should be set to NULL.
Care should be taken to make sure that the color and depth buffer formats are consistent to the formats expected by IceT .Mismatched formats will result in garbage images and possible memory faults.
Also note that when compositing with color blending (icetCompositeMode is set to ICET_COMPOSITE_MODE_BLEND), the color buffer must be rendered with a black background in order for the composite to complete correctly. A colored background can later be added using the background_color as described below.
valid_pixels_viewport is an optional argument that makes it possible to specify a subset of pixels that are valid. In parallel rendering it is common for a single process to render geometry in only a small portion of the image, and IceT can take advantage of this information. If the rendering system identifies such a region, it can be specified with valid_pixels_viewport.
Like all viewports in IceT ,valid_pixels_viewport is an array of 4 integers in the form $<x, y, width, height >$.This viewport is given in relation to the image passed in the color_buffer and depth_buffer arguments. Everything outside of this rectangular region will be ignored. For example, if the valid_pixels_viewport $<10, 20, 150, 100 >$is given, then icetCompositeImage will ignore all pixels in the bottom 10 rows, the left 20 columns, anything above the $160^th$ (10+150) row, and anything to the right of the $120^th$ (20+100) column.
If valid_pixels_viewport is NULL, then all pixels in the input image are assumed to be valid.
projection_matrix and modelview_matrix are optional arguments that specify the projection that was used during rendering. When applied to the geometry bounds information given with icetBoundingBox or icetBoundingVertices, this provides IceT with further information on local image projections. If the given matrices are not the same used in the rendering or the given bounds do not contain the geometry, IceT may clip the geometry in surprising ways. If these arguments are set to NULL, then geometry projection will not be considered when determining what parts of images are valid.
The background_color argument specifies the desired background color for the image. It is given as an array of 4 floating point values specifying, in order, the red, green, blue, and alpha channels of the color in the range from 0.0 to 1.0.
When rendering using a depth buffer, the background color is used to fill in empty regions of images. When rendering using color blending, the background color is used to correct colored backgrounds.
As stated previously, color blended compositing only works correctly if the images are rendered with a clear black background. Otherwise the background color will be added multiple times by each process that contains geometry in that pixel. If the ICET_CORRECT_COLORED_BACKGROUND feature is enabled, this background color is blended back into the final composited image.
On each .igdisplay processdisplay process (as defined by icetAddTile), icetCompositeImage returns the fully composited image in an IceTImage object. The contents of the image are undefined for any non-display process.
If the ICET_COMPOSITE_ONE_BUFFER option is on and both a color and depth buffer is specified with icetSetColorFormatand icetSetDepthFormat,then the returned image might be missing the depth buffer. The rational behind this option is that often both the color and depth buffer is necessary in order to composite the color buffer, but the composited depth buffer is not needed. In this case, the compositing might save some time by not transferring depth information at the latter stage of compositing.
The returned image uses memory buffers that will be reclaimed the next time IceT renders or composites a frame. Do not use this image after the next call to icetCompositeImage (unless you have changed the IceT context).
An argument is set to NULL where data is required.
Not enough memory left to hold intermittent frame buffers and other temporary data.
icetDrawFrame may also indirectly raise an error if there is an issue with the strategy or callback.
An argument to icetCompositeImage is inconsistent with the current IceT state.
The images provided must match the format expected by IceT or else unpredictable behavior may occur. The images must also be carefully rendered to follow the provided viewport and projections. Images that a color blended must be rendered with a black background and rendered with the correct alpha value.
If compositing with color blending on, the image returned may have a black background instead of the background_color requested. This can be corrected by blending the returned image over the desired background. This will be done for you if the ICET_CORRECT_COLORED_BACKGROUND feature is enabled.
Copyright (C)2014 Sandia Corporation
Under the terms of Contract DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains certain rights in this software.
This source code is released under the New BSD License.