Imagine you want to mask/hide some parts of your images, like windows, water, sky, shadows, vegetation, or noise, so that they will not be used in the reconstruction. Masks can also help the reconstruction when you need to turn the scanned object on its side, upside down or just around to capture all details in a more convenient way. You can create alpha masks in an external image editing software and bring the masked images into RealityCapture (RC) or generate and export them automatically directly through this application.
NOTE: Using masks with RealityCapture is not necessary, but can improve the results and speed up the process when all else fails, so its use is rather specific to your particular case. Using masks may help you to filter-out unwanted geometry during the reconstruction phase and hence simplify further model cleaning, e.g., by using lasso and other filtering tools.
Masks represent a non-destructive way of editing images. They may be separate (.png) images or the information on your selection can be added to the originals (saved as tiffs).They are stored in alpha channels and come as grayscale images.
Areas painted black are unmasked, the white areas are masked. You can possibly use up to 256 shades of grey in an alpha channel. However, for purposes of use in RC, we do not recommend you to use any level of transparency in the masked areas of images and apply any filter in the images, since it could confuse its algorithms during processing. The alpha channels are specifically used during feature detection and meshing.
Process of image masks creation:
- Import first set of photos
- Align the images
- Adjust the reconstruction region and generate the preview mesh
- Filter unwanted parts
- Export the masks by Reconstruction tab -> Export -> Depth and Mask
- Save the masks next to the original images
- Process other loops of the objects same way, you can use for every loop new project or disable unwanted cameras, when you work with more loops
- Create new project in RealityCapture
- Import all folders with images and their masks to the project
- Align the images, adjust ground and do the reconstruction
Image masking tutorial video.