Segmentation mask

As explained in Geometry for Programmers

An image mask is a 3D image whose cells indicate, typically with a single bit, whether corresponding spatial regions belong to a selected object or structure.

A segmentation mask is an equal-size 3D image that replaces grayscale measurements with a bit indicating whether each voxel belongs to a selected object or structure. Creating the mask from image data is called segmentation. Masks are also called voxel models. section 12.0

The representation turns a difficult-to-operate grayscale image into a practical spatial model. Filled and unfilled voxels support Boolean intersection, unification, and subtraction, as well as dilation and erosion. section 12.0 section 12.3.4

A mask can support several operations and extensions:

• Intersection keeps voxels filled in both models. section 12.3.4
• Unification keeps voxels filled in either model. section 12.3.4
• Subtraction keeps filled voxels from the first model that the second does not cover. section 12.3.4
• An extended mask can store a distance to the nearest region of interest instead of only a membership bit. section 12.0

Combining voxel models

To combine two segmented bodies, apply Boolean operations directly to their filled and unfilled voxels. This simplicity makes multistep voxel workflows comparatively robust. section 12.3.4

Figure 12.1 The same data (a) represented as an array of colors: a 2D 4×2 image made of pixels (b) and a 3D 2×2×2 image made of voxels (c)

Compare pixels and voxels as arrays: a 3D image is an array of 2D images whose colored cubes represent spatial cells. section 12.0

How Geometry for Programmers covers it

How deep it goes

Defines voxel membership and outlines Boolean operations, while deferring segmentation details and extended mask information.

Examples

Uses filled voxel models, Boolean intersection, unification, subtraction, and distance-valued mask extensions.

What it assumes

Assumes familiarity with 3D images and grayscale data while explaining pixels, voxels, masks, and segmentation from their basic relationship.

What you can do after

Apply Boolean intersection, unification, and subtraction to voxel models and combine them with dilation or erosion.

How it connects

Connects segmentation masks with 3D images, voxels, segmentation, Boolean operations, dilation, and erosion.

liveBookmanning.com

Read Geometry for Programmers on liveBook Get the book at manning.com

Also explained in

• 

  liveBookmanning.com
  Deep Learning with R, Third EditionA segmentation mask is a spatial label representation identifying the pixels belonging to an object or region in an image.
• 

  liveBookmanning.com
  Deep Learning with Python, Third EditionAn image-shaped target annotation for segmentation where each pixel stores the class or region of the corresponding input pixel.
• 

  liveBookmanning.com
  Deep Learning with PyTorch, Second EditionA labeled image-like boolean map indicating which pixels belong to the target region in segmentation.

Compare how 4 books explain segmentation mask →

Related concepts

• 3D imagekind of · A 3D image is a three-dimensional sampled representation of geometry or spatial data arranged as image-like values.
• SAMused by · SAM, or Segment Anything Model, is a pretrained promptable computer-vision model that generates segmentation masks for objects in images.
• Intersection over Unionmeasured by · Intersection over Union is the ratio of the intersection area or volume of two regions to the area or volume of their union.
• voxelequivalent to · A voxel is a discrete volumetric element, typically cubical, used to represent spatial data or occupancy in a three-dimensional grid.
• labelequivalent to · A label is a value or annotation associated with an item that identifies its target category, outcome, or reference answer for learning or evaluation.

This explainer was generated by Manning's knowledge graph from the text of Geometry for Programmers. The section links point to the parts of the book it draws on.