/rawcompr

Losslessly compress multimedia files containing raw video tracks

Primary LanguageC++GNU General Public License v2.0GPL-2.0

rawcompr

Losslessly compress multimedia files containing raw video tracks

Description

This FFmpeg-based tool compresses raw video tracks in multimedia files using one of the following lossless codecs:

  • FFV1

  • HUFFYUV

  • H264 with lossless parameters

The resulting compressed file will be in Matroska video format (.mkv), which can be read by FFmpeg and played by VLC.

Note that while plain simple ffmpeg can encode video tracks into FFV1 (or another lossless codec) and guarantee that every single data bit of the original raw frames can be reconstructed, this tool adds an extra layer that guarantees that the entire original file can be reconstructed (not just single frames one-by-one), including the exact container structure and metadata embedded in the file.

The decompression routine is also implemented in this tool, along with hash-based checks to validate the integrity of the reconstructed file.

How does it work?

Compressing using this tool is roughly equivalent to

ffmpeg -i INPUT_FILE.avi -codec:v ffv1 -codec:a copy OUTPUT_FILE.mkv

with the difference that it will also output an extra file, with .llr extension ("LossLess References"), which is basically a copy of the original file, with two differences:

  • sections originally containing frames' raw data are replaced with references to the corresponding video frames in the .mkv file

  • similarly, sections originally containing other types of frames (e.g. audio frames) are also replaced with references to the corresponding frames (e.g. audio frames) in the .mkv file.

In addition, the .llr file contains metadata such as the hash of the original file.

The decompressor reverses the process by copying the original file’s contents from the .llr file, while replacing references with:

  • a decompressed video frame (wherever the original file had a raw video frame)

  • a copy of the corresponding .mkv frame (wherever the original file had a different type of frame)

Lastly, at the end of the process, the decompression routine compares the hash of the reconstructed file to the hash contained in the .llr metadata section to validate the result.

Usage

$ rawcompr -h
Losslessly compress raw streams in multimedia files.

Usage: rawcompr [-d] [OTHER OPTIONS] -i INPUT OUTPUT

Basic options:
 -d        Decompress instead of compressing
 -i INPUT  Input file
 OUTPUT    Output file
 --debug   Enable debug output from rawcompr
 --libavloglevel LEVEL
           Set libav log level

Compression-only parameters:
 -v CODEC_NAME [key=value ...]
           Select video codec and options
 --hash ALGORITHM
           Embed the input file's hash using the selected algorithm (default: MD5)

Note:
 - If compressing, OUTPUT file must have .mkv extension
 - If decompressing, INPUT file must have .mkv extension

[cut]

rawcompr -h also shows the default codec (FFV1) and options, and the list of available hashing algorithms.

Compression and decompression

$ md5sum original.avi
569214dc03ca67ec4007860851df9fee  original.avi

$ rawcompr -i original.avi compressed.mkv && rm original.avi

$ ls
compressed.llr  compressed.mkv

$ rawcompr -d -i compressed.mkv reconstructed.avi

$ md5sum reconstructed.avi
569214dc03ca67ec4007860851df9fee  reconstructed.avi

Note: The original file can be in any multimedia format, not just AVI.

Note 2: the two md5sum invocations were listed for clarity’s sake. Hash verification is already built-in in the decompression algorithm.

Supported codecs (with tested options) and comparison

  • FFV1 (default options, see rawcompr -h): rawcompr -i original.avi compressed-default.mkv

  • FFV1 (no options): rawcompr -i original.avi -v ffv1 compressed-ffv1.mkv

  • HUFFYUV: rawcompr -i original.avi -v huffyuv compressed-huffyuv.mkv

  • Lossless H264: rawcompr -i original.avi -v h264 crf=0 tune=zerolatency compressed-h264.mkv

Note: With H264, crf=0 is necessary to setup lossless compression and tune=zerolatency is necessary due to how rawcompr interacts libav.

File Name File Size Notes

original.avi

2.2 GiB

720x576 YUY2 video + PCM stereo audio, duration: 113 seconds

compressed-default.mkv

850.0 MiB

compression: 78.4 seconds, decompression: 94.5 seconds

compressed-ffv1.mkv

948.4 MiB

compression: 52.5 seconds, decompression: 54.9 seconds

compressed-huffyuv.mkv

1.1 GiB

compression: 11.7 seconds, decompression: 15.2 seconds

compressed-h264.mkv

965.8 MiB

compression: 71.9 seconds, decompression: 89.1 seconds

Compression and decompression times were measured on an Intel i7-4910MQ CPU.

Installation

Prerequisites:

  • Fedora (with RPMfusion free repository): dnf install cmake gcc-c++ ffmpeg-devel

  • Debian/Ubuntu: apt install cmake g++ libavcodec-dev libavformat-dev libswscale-dev make pkg-config

Build and install:

mkdir build
cd build
cmake ..
make
sudo make install