Audio/Video Quality Testing: Using ffmpeg Filters and VQI for Black Screen, Freeze, Blur, and Silence Detection

Tool Introduction

ffmpeg provides a rich set of filters (ffmpeg‑filters) that can modify or analyse raw audio/video frames, such as scaling, cropping, flipping, and overlaying. Filters can be chained into a filtergraph to perform complex processing.

VQI (Video Quality Indicators) is an open‑source suite that can evaluate various video quality metrics (black screen, block loss, blur, freezing, letter‑boxing, etc.) on macOS, Windows, and Linux, with scripts available for bash, batch, Python, and MATLAB.

Black Screen Detection

ffmpeg blackdetect filter

Detects periods where the frame is considered black based on configurable thresholds.

Key options:

black_min_duration (d): minimum duration of a black segment (default 2.0 s).

picture_black_ratio_th (pic_th): proportion of black pixels in a frame (default 0.98).

pixel_black_th (pix_th): pixel‑level black threshold (default 0.10).

Example command:

ffmpeg -i INPUT -vf blackdetect=d=2:pix_th=0.00:pic_th=0.9 OUTPUT

ffmpeg blackframe filter

Provides detailed information per black frame, including frame number, black pixel ratio, PTS, and time.

Key options:

threshold (thresh): pixel black threshold (default 32).

amount: percentage of black pixels required to flag a frame (default 98).

VQI black frame detection

Outputs a simple CSV where each frame is marked 1 for black frame and 0 otherwise.

Example usage:

sh mitsuScript.sh <mp4_dir> <path_of_vqibinary>

Block‑Loss (Blur) Detection

Block‑loss appears as irregular colored pixel blocks caused by stream corruption. VQI’s blockloss indicator reports a value from 0 to 5 (higher means more severe).

VQI’s Blur indicator similarly ranges 0‑5, with larger values indicating greater blur.

Freezing (Stutter) Detection

Frames that linger longer than a threshold cause visible stutter. VQI’s Freezing indicator marks such frames with a value of 1, otherwise 0.

Black Border Detection

Black borders (letter‑boxing or pillar‑boxing) may result from mismatched source and output resolutions. VQI provides Letter‑boxing and pillar‑boxing metrics where 0 means no border and larger values indicate larger black areas.

Silence Detection

ffmpeg’s silencedetect audio filter logs silent intervals based on configurable noise tolerance and minimum duration.

Key options:

noise (n): noise tolerance (default –60 dB).

duration (d): minimum silence length to report (default 2 s).

Example command:

ffmpeg -i INPUT -af silencedetect=d=5:n=-50dB OUTPUT

Volume Check

ffmpeg’s volumedetect filter reports RMS and maximum volume levels in dB.

Example command:

ffmpeg -i INPUT -filter_complex volumedetect -c:v copy -f null /dev/null

These tools together enable comprehensive audio/video quality testing beyond basic playback verification.