How to Use the FFmpeg mestimate Filter

The FFmpeg mestimate (motion estimation) filter is a powerful tool used to estimate and reconstruct motion vectors between video frames. This article provides a quick overview and a straightforward guide on how to use the mestimate filter, explaining its key parameters, basic syntax, and how to visualize the estimated motion vectors in your video processing workflows.

Understanding the mestimate Filter

The mestimate filter calculates motion vectors by comparing macroblocks between consecutive frames. Rather than just using the motion vectors already encoded in a compressed video stream, mestimate allows you to force a recalculation of motion using specific search algorithms. This is highly useful for video analysis, custom encoding preparations, and research.

Basic Syntax

The basic syntax for the mestimate filter is:

ffmpeg -i input.mp4 -vf "mestimate=method=esa:mb_size=16:search_param=7" output.mp4

Key Parameters of mestimate

You can customize the behavior of the motion estimation process using the following parameters:

• method: Specifies the motion estimation search method. Available options include:
  • esa: Exhaustive search algorithm (slowest, most accurate).
  • tss: Three-step search algorithm.
  • tdls: Two-dimensional logarithmic search algorithm.
  • ntss: New three-step search algorithm.
  • fss: Four-step search algorithm.
  • ds: Diamond search algorithm (fast).
  • hex: Hexagon-based search algorithm.
  • epzs: Enhanced predictive zonal search algorithm (default).
  • umh: Uneven multi-hexagon search algorithm.
• mb_size: Sets the macroblock size. The default value is 16 (16x16 pixels).
• search_param: Sets the search parameter radius. The default value is 7.

Practical Examples

1. Basic Motion Estimation

To apply motion estimation using the default EPZS method on an input video, run:

ffmpeg -i input.mp4 -vf "mestimate" output.mp4

2. Using Exhaustive Search for High Accuracy

If you require high-accuracy motion estimation and do not mind longer processing times, use the Exhaustive Search Algorithm (esa):

ffmpeg -i input.mp4 -vf "mestimate=method=esa:mb_size=16:search_param=15" output.mp4

3. Visualizing Motion Vectors

The mestimate filter calculates motion vectors and attaches them as side data to the video frames. To actually see these vectors drawn on top of your video, you must pair mestimate with the codecview filter.

Run the following command to calculate and visualize forward predicted motion vectors:

ffmpeg -i input.mp4 -vf "mestimate,codecview=mv=pf" output.mp4

• mv=pf: Visualizes forward predicted P-frame motion vectors.
• mv=bf: Visualizes forward predicted B-frame motion vectors.
• mv=bb: Visualizes backward predicted B-frame motion vectors.

You can combine them to visualize all vectors at once:

ffmpeg -i input.mp4 -vf "mestimate,codecview=mv=pf+bf+bb" output.mp4