Skip to content

Face detection and processing pipeline

Updated 11 Oct 2026

•
Face detection
- •
  Detection locates likely face regions; it does not identify whose face it is.
- •
  Haar cascades scan image windows for learned patterns; results vary with pose, illumination, scale, and occlusion.
- •
  YuNet -> another detector supported through OpenCV.
- •
  Detection output commonly includes bounding boxes and scores that need thresholding and visualization.
•
Project pipeline
- •
  Request flow
  - •
    FastAPI accepts an uploaded image.
  - •
    OpenCV decodes and processes it.
  - •
    The application returns or displays the result.
- •
  Operation order -> each step receives the previous step’s output.
  - •
    noise addition order and filtering change later inputs
- •
  Residual -> input minus low-pass image.
- •
  "Band pass" branch -> adds a low-pass image to that residual
  - •
    result -> approximately reconstructs the original; does not select a distinct frequency band.
- •
  “High pass” branch -> applies sharpening
- •
  Frequency-noise option -> zeros a central low-frequency square.
  - •
    Effect -> suppresses low frequencies.
  - •
    It does not target periodic peak pairs like a notch filter
- •
  app-flow diagram: upload -> FastAPI -> OpenCV decode/process -> result -> browser display; show where each filter option runs.

Linked references 1

digital-image-processing-polban

◌ Explore connections in Graph view