CV MCQ — Chapter 3 0 Questions
Segmentation Basics

Segmentation Basics MCQ

Thresholding (global, Otsu, adaptive) and morphological operations for binary masks and region cleanup.

Easy: 0 Q Medium: 0 Q Hard: 0 Q

Image Thresholding MCQ

Thresholding for segmentation

Thresholding partitions pixels into foreground and background (or multiple levels). It is fast but assumes separable intensity distributions.

Otsu’s method

Chooses T by maximizing between-class variance for a bimodal-ish histogram—automatic global threshold when classes are separable.

When to use what

Global

One T for the whole image—simple, fails under uneven illumination.

Adaptive

Local mean/Gaussian-weighted thresholds per neighborhood—handles shading gradients.

Invert & polarity

Know whether objects are dark-on-bright or bright-on-dark; invert maps if needed.

Post-process

Morphology often cleans threshold noise (salt-and-pepper) before contour extraction.

Pipeline snippet

Optional blur → Threshold → Morphology → Contours / features

Pro tip: Inspect the histogram before thresholding—unimodal scenes rarely binarize cleanly with one T.

Morphological Operations MCQ

Mathematical morphology

Morphology processes shapes using a structuring element (SE): erosion shrinks foreground, dilation grows it; openings remove speckles, closings fill gaps.

Duality

Erosion on foreground relates to dilation on background—design SE size/shape for the defect scale you target.

Operators

Erosion

Local minimum over SE support—removes thin protrusions, separates touching objects if SE fits.

Dilation

Local maximum—bridges small gaps, thickens strokes.

Opening

Erosion then dilation—suppresses bright noise smaller than SE.

Closing

Dilation then erosion—fills dark holes smaller than SE.

Post-threshold cleanup

Binary map → Opening/Closing → Connected components

Pro tip: SE anisotropy matters—lines along one direction erode faster with orthogonal SEs.