More info here
Some tutorials for the compositing nodes can be found here.
Items below marked as (fast) have been rewritten to run about 5 times faster than before. The mask portion of these nodes if any still uses older slower techniques.
Input a grayscale bump map to generate a normal map.
Uses it's animation keys to create a curve for modifying the values of an image. Used as an input to the image curve node. Open the FCurve animation tab to make changes to the curve. Create keys between frame0 and 100 with values from 0 to 1. Zoom in to fit this time and value range to the view. Push the "Start Interactive Update" to see a live update of the changes. Push stop to end the interactive update. Response can be slow while in interactive update mode. Use the "image shrink and zoom" node to reduce the resolution of the image and the period for the update can be increased.
Choose/Examine a color from an image or use as a preview display.
Option to drag and drop an image onto the thumbnail or ctrl click to use a file dialog
Save swatch button for use with the web based colorpicker.
Add the values of two images.
Combine images by alternating lines. Value of 2 swaps images every 2 lines. 1 swaps every other line.
Create an image based on the maximum rgba values of two images.
Create an image based on the minimum rgba values of two images.
Create an image based on the weighted rgba values of two images. A mix value of 0.5 will mix equal amounts of the 2 images. The mix value is between 0 and 1. The mix amount can also be driven by an input image. Check mix alpha to read the mix value from the image alpha channel, uncheck to read from the color portion of the image.
Value = bitmapA * mix + bitmapB * (1-mix)
Create an image based on the multiplication of two images. The Divide checkbox will change the node to divide instead of multiply.
Display the composite result through a camera on an image plane. The disnatncein controls how far the image plane is from the camera and the xsize and ysize control the image aspect ratio. To use select a camera and push the "Connect to Camera" button then open a 3d view and set it to camera view mode. To fit the image in the view also use the window settings to set the width and height to match. A second version of this node is included that takes an alpha bitmap input.
Combine two images with the composite over process using the alpha channel of bitmapA or if connected the "Input alpha" image. The "alpha alpha" check will take the alpha value from the alpha channel of the "Input alpha" image. Otherwise it will use the color of the "Input alpha" image to control the alpha.
Combine two images using overlay mode. Reverse the input connections for a Hard Light mode.
Combine two images using screen mode.
Combine two images using subtraction. Use absolute value to use as a difference.
Add a constant value to an image. The "alpha value only" will add the "Alpha In" constant to each of the rgb values of the image, otherwise the rgba constants will be added to the corresponding rgba values in the input image. A mask input image can be used to apply the change to portions of the input. Mask alpha only will use the mask alpha value instead of the rgb values of the mask.
Blur an image. A mask input image can be used to apply the change to portions of the input. Mask alpha only will use the mask alpha instead of the rgb values of the mask.
Blur code taken from http://www.blackpawn.com/texts/blur/default.html
Combine 4 input images to a single rgba image. All inputs must be used to generate an output. Only the red channel of each input image will be read. Gray images in truespace use the red channel.
Split an image into it's color components, rgba.
Computes the distance from the input color defined by the rgb values and generates an image based on the color distance.
Basic green/blue screen image processor. Uncheck for blue screen.
This convolve has the effect of sharpening the image. Check edge finder to output the image lines. Mask input is used to restrict the effect to areas defined by it's rgb or alpha value. Radius has no effect and is hard coded to 1 inside the node.
Input a color curve to control the output effect on an image. Alpha Curve Only checked means to only look at the alpha color curve. If unchecked each of the rgba color curves will be used.
Apply gamma to an input image with optional mask and optional separate rgb gammas.
Invert the input image.
Load a single image or sequence of images. Check "isaSequence" to load a sequence. Push the "setup keyframes" button to set keyframes needed to switch images as the timeline progresses.
An experimental/unfinished node called "image sequence loader with scale" is included in the library/zip. Do not use it.
Convert rgb to grayscale image.
Multiply an image by a constant. Alpha Value Only check will multiply the rgb by the alpha in input. Unchecked and each rgba will be multiplied individually. Any contrast amount other than 1 will apply a contrast to the image. Option mask input to restrict the effect to a mask image values.
Perlin noise. X and Y Scale controls the size of the noise pattern. X and Y Size controls the image resolution
based on Blender code
Store an image rgb values to the alpha channel. Use the check box to reverse and store the alpha to the rgb values.
Save the result to disk. Don't check "enableSave" until all else is setup. Click save as to define the base file name and type of image(choose "all" to change from the tga default). Check "isaSequence" to save a sequence of images. Set the timeline to the desired output range and push the setup keyframes button so that images will be saved as the time progresses. It may be necessary to set playback to all frames(not sure about that). Uncheck repeat in the animation preferences so it only plays once. last step is to rewind the animation, check enableSave and play the animation. In this case the animation would be the sequence of images processed through the compositing system.
The composite render node can also be used to render the sequence from a camera, but alpha values and any hdr values will not be preserved. Extensions must be typed in manually in the save as dialog. Click here for an updated image saver with extensions included.
BMP - Save as bitmap file
JPG - Save as jpeg
TGA - Save as targa
PNG - Save as PNG
DDS - Save as DirectX DDS file
PPM - PPM
DIB - Save as Windows DIB
HDR - Save in HDR format
PFM - Save in Portable Pixmap format
Note(April 2011): Just found that one of my assumptions was wrong. Setting playback to "all frames" does not guarantee that all frames will be evaluated. Use the Truespace "Render to File" to render a dummy sequence which in turn will force each frame to be evaluated.
April 2011 - fixed bug in the image saver number sequencer.
The nodes can be slow. This node will reduce the size of it's input image by resolution divisor factor. 2= half the size in horizontal and vertical, 3= one third the size. The compositing nodes are slow so apply an image shrink to all input images while working. Remove/bypass the image shrink when rendering the final result. Check zoom to zoom the image and X and Y offsets to move the image center(actually the upper left corner). Zoom can be used with the color picker to see a closeup.
Used to create an image of a specific size with a constant color. Select a color then push the button to generate the image.
Used to create a cropped image with the upper left corner at ( X Offset, Y Offset ) and physical dimensions of X and Y Size.
Resolution of X and Y Size.
based on Blender https://docs.blender.org/manual/en/dev/render/blender_render/textures/types/procedural/clouds.html
Resolution of X and Y Size.
based on Blender https://docs.blender.org/manual/en/dev/render/blender_render/textures/types/procedural/marble.html
Resolution X and Y Size.
based on Blender https://docs.blender.org/manual/en/dev/render/blender_render/textures/types/procedural/voronoi.html
Resolution X and Y Size.
based on Blender: https://docs.blender.org/manual/en/dev/render/blender_render/textures/types/procedural/wood.html