How to Apply Spatial-Temporal Deinterlacing in FFmpeg

Deinterlacing is a crucial step when converting legacy interlaced video formats into progressive frames suitable for modern digital displays. This article provides a direct, practical guide on how to use the motion-compensated spatial-temporal deinterlacing filter (mcdeint) in FFmpeg, explaining its syntax, key parameters, and offering ready-to-use command examples alongside modern alternatives.

Understanding the MCDEINT Filter

The mcdeint (motion-compensated deinterlacer) filter is an advanced spatial-temporal deinterlacing filter in FFmpeg. Unlike simple spatial deinterlacers that only look at a single frame, or temporal deinterlacers that simply merge fields, a spatial-temporal filter analyzes both the current frame and surrounding frames in time.

By utilizing motion estimation, mcdeint tracks moving objects across frames to reconstruct missing lines, resulting in much cleaner edges and significantly fewer interlacing artifacts (like “comb” effects) without sacrificing image sharpness.

FFmpeg Syntax for MCDEINT

The basic syntax for applying the mcdeint filter in an FFmpeg command is as follows:

ffmpeg -i input.mp4 -vf "mcdeint=mode:parity:qp" output.mp4

The filter accepts three optional parameters separated by colons:

1. mode: Controls the processing intensity and motion estimation quality.
   • 0: Fast (basic spatial-temporal processing).
   • 1: Medium (better motion estimation).
   • 2: Slow (uses detailed motion estimation).
   • 3: Extra slow (highest quality, highly CPU-intensive).
2. parity: Determines the field dominance (which field comes first).
   • 0: Bottom Field First (BFF).
   • 1: Top Field First (TFF).
   • -1: Auto-detect (recommended).
3. qp: Quantization Parameter. Higher values allow smoother motion search but may reduce spatial sharpness. The default is typically sufficient for most uses.

Practical Command Examples

To apply standard-quality motion-compensated deinterlacing with automatic field parity detection, use this command:

ffmpeg -i input_interlaced.mp4 -vf "mcdeint=mode=1:parity=-1" -c:v libx264 -crf 20 output_progressive.mp4

For the highest possible quality (which requires significant processing time), increase the mode to 3:

ffmpeg -i input_interlaced.mp4 -vf "mcdeint=mode=3:parity=-1" -c:v libx264 -crf 20 output_high_quality.mp4

Important Usage Considerations

While mcdeint produces high-quality results, it has several limitations in modern workflows:

• Compilation Requirements: The mcdeint filter requires FFmpeg to be compiled with specific legacy components (specifically the Snow codec and libpostproc). In many modern pre-compiled FFmpeg distributions, mcdeint is disabled or unavailable.
• Performance: It is extremely slow and single-threaded, making it impractical for high-definition (HD or 4K) video processing.

Modern Spatial-Temporal Alternatives

If mcdeint is not supported by your FFmpeg build, or if it runs too slowly, you should use modern, highly-optimized spatial-temporal deinterlacers:

1. YADIF (Yet Another Deinterlacing Filter)

YADIF is the industry standard for fast, high-quality spatial-temporal deinterlacing.

ffmpeg -i input_interlaced.mp4 -vf "yadif=mode=0:parity=-1" -c:v libx264 output.mp4

(Note: mode=0 outputs one frame per frame of video. mode=1 outputs one frame per field, creating a 60fps progressive video from a 30i interlaced source).

2. BWDIF (Bob Weaver Deinterlacing Filter)

BWDIF is a newer, hardware-optimized spatial-temporal filter based on YADIF algorithms that delivers better detail preservation.

ffmpeg -i input_interlaced.mp4 -vf "bwdif=mode=0:parity=-1" -c:v libx264 output.mp4