Creating Holdouts with the DeepMerge Node
The holdout operation of the DeepMerge node removes or fades out samples in input B that are occluded by samples in input A.
To create a holdout using the DeepMerge node:
1. | Connect the deep image you want to remove or fade parts from to input B. |
2. | Connect the deep image with the occluding parts to input A. |
3. | In the DeepMerge properties, set operation to holdout. |
4. | You can filter out samples using the drop zero threshold control. Increasing the value removes more samples with very small alpha values, such as those caused by floating point inaccuracy. |
5. | The metadata from control allows you to control which input's metadata is passed down the node tree. |
Note: When metadata from is set to All and there are keys with the same name in both inputs, keys in B override keys in A.
6. | Enable volumetric holdout if you want Nuke to calculate occlusion using the values of the holdout samples in front of samples from main. This is a more accurate representation of occlusion at depth, but can take longer to process. |
For example:
M = main sample
H = holdout sample
M0 remains unchanged since there are no holdout samples before it. M2 is affected by the combined H0, H1, and H2 holdout samples and M4 is affected by all holdout samples.
You can now view the result, which is a holdout with red, green, blue, and alpha channels. Note that the output image is still a deep image.
Note: When volumetric holdout is disabled, deep samples that coincide with a holdout are removed and everything 'deeper' is also removed.