Computer Vision Interview 20 essential Q&A Updated 2026
filtering

Image Filtering: 20 Essential Q&A

Linear filtering, convolution vs correlation, blur types, and how kernels tie to low-pass and high-pass ideas.

~11 min read 20 questions Beginner–Intermediate
convolutionGaussianmedianpadding
Quick Navigation
1. What is image filtering?
2. Convolution definition
3. Correlation vs convolution
4. Kernel / mask
5. Linear vs nonlinear filters
6. Gaussian filter
7. Separable Gaussian
8. Box / mean filter
9. Median filter
10. Bilateral filter (idea)
11. Padding modes
12. Border handling
13. Sharpening kernels
14. Unsharp mask
15. Stride in CNNs vs spatial filtering
16. Frequency intuition
17. Noise removal tradeoffs
18. Derivative filters preview
19. Kernel normalization
20. OpenCV filter2D / blur
1 What is image filtering? ⚡ easy
Answer: Computing a new image where each output pixel is a function of a neighborhood of input pixels. Linear filters use weighted sums (convolution/correlation); nonlinear filters include median, bilateral, morphological ops.
2 Define 2D convolution (discrete). 📊 medium
Answer: Slide a kernel over the image; at each location, sum of elementwise products of kernel and flipped neighborhood (strict convolution). Many libraries implement cross-correlation without flip—know which convention your framework uses.
3 Correlation vs convolution for symmetric kernels? 📊 medium
Answer: For symmetric kernels (Gaussian), results match. For asymmetric kernels (Sobel direction), flip matters for strict signal-processing convolution vs correlation.
4 What is a kernel / mask? ⚡ easy
Answer: Small matrix of weights defining neighborhood contributions. Size (e.g. 3×3, 5×5) sets spatial support; larger kernels increase blur radius and compute cost (~kernel area per pixel).
5 Examples of nonlinear filters? ⚡ easy
Answer: Median (order-statistic, good for salt-and-pepper), bilateral (edge-preserving smoothing), morphology (min/max). They do not obey superposition like convolutions.
6 Why use a Gaussian kernel? 📊 medium
Answer: Smooth low-pass filtering that reduces noise and high frequencies while avoiding sharp ringing like ideal low-pass. Separable implementation makes it fast; σ controls blur strength. 
import cv2
blur = cv2.GaussianBlur(img, (5, 5), sigmaX=1.0)
7 What does separable filter mean? 🔥 hard
Answer: A 2D kernel K can equal outer product v vᵀ. Convolving with K is equivalent to 1D conv along rows then columns—cost drops from O(WHk²) to O(2WHk) for k×k support.
8 Mean / box filter properties? ⚡ easy
Answer: Simple average of neighborhood—fast (especially with integral images) but has a sharp frequency nulls profile vs Gaussian; can create blocky artifacts compared to Gaussian blur.
9 When prefer median filtering? 📊 medium
Answer: Impulsive salt-and-pepper noise where mean blur smears outliers. Median preserves edges better than Gaussian for that noise but is costlier and can remove thin structures.
10 Intuition for bilateral filter? 🔥 hard
Answer: Weighted average where weights drop with both spatial distance and intensity difference—smooths flat regions but preserves sharp edges. Used for denoising and tone mapping; slower than Gaussian.
11 What is padding in convolution? 📊 medium
Answer: Extends the image border so output size can match input (same padding) or follow strict convolution (valid). Modes: zero, reflect, replicate, wrap—choice affects edges and CNN behavior.
12 Why do edges look different after filtering? ⚡ easy
Answer: Neighborhoods at borders are incomplete; padding synthesizes missing neighbors. Wrong padding can cause dark/bright fringes—noticeable on small images and CNN feature maps.
13 Basic sharpening idea? 📊 medium
Answer: Emphasize high frequencies by adding a scaled Laplacian-like response or subtracting a blurred version from the original—makes edges pop but can amplify noise.
14 What is unsharp masking? 📊 medium
Answer: Enhancement: original + amount × (original − blurred). The difference is a high-boost of details; used in photography and preprocessing (with care for noise).
15 How is CNN stride related to classical filtering? 📊 medium
Answer: Stride >1 subsamples the output—like convolve-then-downsample. Larger stride increases receptive field progression and reduces spatial size; different from stride-1 spatial filtering used in preprocessing.
16 Frequency view of Gaussian blur? 🔥 hard
Answer: Gaussian in space ↔ Gaussian in frequency; it attenuates high frequencies smoothly. Helps before subsampling to limit aliasing (Nyquist)—ties back to image basics.
17 Gaussian noise vs salt-and-pepper—filter choice? 📊 medium
Answer: Gaussian noise: linear smoothing (Gaussian blur) or Wiener/BM3D-class methods at higher level. Salt-and-pepper: median or morphological openings/closings.
18 How do derivative filters relate to filtering? 📊 medium
Answer: Finite differences (Sobel/Prewitt) are short convolution kernels approximating gradients—high-pass. Often paired with prior Gaussian smoothing to reduce noise before edge detection.
19 Why normalize blur kernels? ⚡ easy
Answer: So the DC gain is 1—preserves average brightness. Unnormalized Gaussian sums to 1 after discretization normalization; forgetting normalization scales image intensity.
20 OpenCV: GaussianBlur vs filter2D? ⚡ easy
Answer: GaussianBlur builds separable Gaussian internally. filter2D applies arbitrary kernel (correlation-style in OpenCV)—flexible for custom linear filters.

Filtering Cheat Sheet

Linear
  • Convolution / correlation
  • Gaussian low-pass
  • Separability
Nonlinear
  • Median (impulse noise)
  • Bilateral (edge-aware)
  • Morphology (later topic)
CNN tie-in
  • Learned kernels
  • Padding & stride
  • Depthwise convs

💡 Pro tip: Mention separability + complexity when discussing large Gaussian kernels.

Full tutorial track

Go deeper with the matching tutorial chapter and code examples.

Filtering Tutorial
Previous Next