Technology / Restoration

Why "Foggy" Film Isn't Fog — Classical Restoration Done Right

Most "foggy" high-school football film suffers from bitrate starvation, crushed blacks, and flat levels rather than literal atmospheric fog.

SECTION

Diagnosis

Calibrated on real football frames, we measure:

  • Bits per pixel (strongest discriminator, 17× range)
  • Full dynamic-range span (YMAX - YMIN)
  • Percentile band (turf uniformity, not flatness)

Blockdetect and blurdetect were measured across clean, synthetically degraded, and real football film. They failed to discriminate (blockdetect 2.826 clean vs 2.811 degraded) and were removed from the pipeline.

SECTION

The Classical Ladder (applied in measured order)

  1. Black-level lift to correct crushed shadows.
  2. Contrast stretch driven by occupied range (skipped if span > 0.88 — typical for good football film at ~0.91).
  3. Content-aware denoising.
  4. Sharpening.
CLASSICAL RESTORATION — BEFORE / AFTER
Football film restoration before and after
BEFORE — BITRATE STARVED
AFTER — CLEANED FOR STUDY
CLASSICAL LADDER APPLIED • NO GENERATIVE INVENTION

Real high-school football film. The enhanced viewing derivative (right) is for coaches only. Training data always uses the immutable master (left).

The output enhanced video is provided for coaches to watch. It is deliberately never used as training data. Generative "super-resolution" invents detail that does not exist in the original 6-pixel-tall players; training on those hallucinations would teach the model the generative prior instead of real football physics and appearance (core Decision D2).

All computer-vision labels and crops always resolve against the immutable master file.