Compare Videos with FFmpeg maskedmin Filter

This article explains how to use the maskedmin filter in FFmpeg to compare two video streams and output the minimum pixel values. You will learn the basic concept of how this three-input filter works, the required command-line syntax, and a practical example to implement it in your video processing projects.

Understanding the maskedmin Filter

The maskedmin filter is a pixel-level comparison filter in FFmpeg. Unlike simple two-stream comparison filters, maskedmin requires three video inputs of identical resolutions and pixel formats:

1. Primary Stream (Source 1): The base video stream.
2. Reference Stream (Source 2): The video stream compared against the primary stream.
3. Mask Stream: The stream that controls where the comparison takes place.

For each pixel, the filter checks the value of the mask stream. If the mask pixel value is high (usually above a default threshold), the filter outputs the minimum pixel value between the Primary and Reference streams. If the mask pixel value is low, the filter outputs the pixel value from the Primary stream.

Basic Command Syntax

To use the maskedmin filter, you must pass three video inputs to -filter_complex and map them to the filter. Here is the standard command-line template:

ffmpeg -i video1.mp4 -i video2.mp4 -i mask.mp4 -filter_complex "[0:v][1:v][2:v]maskedmin=planes=15[out]" -map "[out]" output.mp4

Parameter Breakdown

• [0:v][1:v][2:v]: Specifies the three input streams in order (Primary, Reference, and Mask).
• maskedmin: Calls the filter.
• planes: Specifies which color planes to process. The value is a bitmask (e.g., 15 processes all planes: Y, U, V, and Alpha). If omitted, the filter defaults to processing all available planes.

Practical Example

If you want to compare two identical-sized videos and use a static grayscale image as a mask (to restrict the minimum-value comparison to a specific region of the frame), you can loop the image as the third input:

ffmpeg -i main_video.mp4 -i ref_video.mp4 -loop 1 -i mask_image.png -filter_complex "[0:v][1:v][2:v]maskedmin[out]" -map "[out]" -shortest output.mp4

In this setup, the minimum pixel values between main_video.mp4 and ref_video.mp4 will only be rendered in the output where the mask_image.png is white or bright. Where the mask image is black, the output will default to the pixels of main_video.mp4.