Membrane FFmpeg SWScale plugin

Plugin providing an element scaling raw video frames and performing pixel format conversions, using SWScale module of FFmpeg library.

It is a part of Membrane Multimedia Framework.

Installation

Add the following line to your deps in mix.exs. Run mix deps.get.

{:membrane_ffmpeg_swscale_plugin, "~> 0.15.0"}

The precompiled builds of the ffmpeg will be pulled and linked automatically. However, should there be any problems, consider installing it manually.

Manual instalation of dependencies

macOS

brew install ffmpeg

Ubuntu

sudo apt-get install ffmpeg

Arch / Manjaro

pacman -S ffmpeg

Description

PixelFormatConverter

PixelFormatConverter accepts raw video in any of the pixel formats specified in type Membrane.RawVideo.pixel_format_t() as input and is capable of producing output in any of these pixel formats.

When creating the element you need to specify a single option format defining the desired pixel format of the output. The element requires Membrane.RawVideo stream format on the input with aligned: true constraint, meaning that each buffer must contain exactly one raw video frame.

Scaler

Scaler needs input in the YUV420p format, processes one frame at a time and requires getting stream format with input video width and height. To meet all requirements either Membrane.RawVideo.Parser or some decoder (e.g. Membrane.H264.FFmpeg.Decoder) have to precede Scaler in the pipeline.

There are two options that have to be specified when creating the element:

output_width - desired scaled video width.
output_height - desired scaled video height.

Both need to be even numbers.

Scaling consists of two operations:

scaling itself - resizing video frame with keeping original ratio. After that operation at least one of the dimensions of the input frame match the respective dimension of the desired output size. The second one (if does not match) is smaller than its respective dimension.
adding paddings - if one dimension does not match after scaling, paddings have to be added. They are put on both sides of the scaled frame equally. They are either above and below the frame or on the left and right sides of it. It depends on the dimension that did not match after scaling.

The output of the element is also in the YUV420p format. It has the size as specified in the options.

Usage

`Membrane.FFmpeg.SWScale.Scaler`

Scaling an encoded (using H.264 standard) video requires parser and decoder because Scaler scales only raw, decoded video. The pipeline scales the video and reencodes it.

defmodule Scaling.Pipeline do
  use Membrane.Pipeline

  @impl true
  def handle_init(_ctx, _options) do
    structure = [
      child(:file_src, %Membrane.File.Source{location: "/tmp/input.h264"})
      |> child(:parser, Membrane.H264.Parser)
      |> child(:decoder, Membrane.H264.FFmpeg.Decoder)
      |> child(:scaler, %Membrane.FFmpeg.SWScale.Scaler{output_width: 640, output_height: 640})
      |> child(:encoder, Membrane.H264.FFmpeg.Encoder)
      |> child(:file_sink, %Membrane.File.Sink{location: "/tmp/output.h264"})
    ]

    {[spec: structure}, %{}}
  end
end

`Membrane.FFmpeg.SWScale.PixelFormatConverter`

Converting the pixel format of an encoded (using H.264 standard) video requires a parser and a decoder because the Converter performs conversion only on the raw , decoded video. The pipeline changes the pixel format of the video to the I422 format and reencodes it.

defmodule Converting.Pipeline do
  use Membrane.Pipeline

  @impl true
  def handle_init(_ctx, _options) do
    structure = [
      child(:file_src, %Membrane.File.Source{location: "/tmp/input.h264"})
      |> child(:parser, Membrane.H264.Parser)
      |> child(:decoder, Membrane.H264.FFmpeg.Decoder)
      |> child(:converter, %Membrane.FFmpeg.SWScale.PixelFormatConverter{format: :I422})
      |> child(:encoder, Membrane.H264.FFmpeg.Encoder)
      |> child(:file_sink, %Membrane.File.Sink{location: "/tmp/output.h264"})
    ]
    
    {[spec: structure], %{}}
  end
end

Copyright and License

Licensed under the Apache License, Version 2.0