README.md

OMAF File Creation tools

Provides guidelines for creation of test vectors for HEVC-based viewport-dependent OMAF video profile with MCTS.

In the following it is explained how to create OMAF compliant files from raw (yuv) ERP source video files.

Destination format:

• Projection: CMP
• Number of tiles: 24 (24 high and 24 low res tiles)
  • High resolution tile size: 768x768
  • Low resolution tile size: 384x384
• Number of 'viewport' streams (hvc2 tracks) = 24
  • Each 'viewport' stream consists of: 12 high resolution tiles and 12 low resolution tiles with following packing:

+-------+--------+-------+--------+---|---+
|       |        |       |        |   |   |
|       |        |       |        ----|----
|       |        |       |        |   |   |
+-------+--------+-------+--------+---|---+
|       |        |       |        |   |   |
|       |        |       |        |---|----
|       |        |       |        |   |   |
+-------+--------+-------+--------+---|---+
|       |        |       |        |   |   |
|       |        |       |        |---|----
|       |        |       |        |   |   |
+-------+--------+-------+--------+---|---+

Decoded (packed) frame:

Rendered (projected) frame:

SRQR regions visualisation:

Required tools

Following tools are required for this script to work:

• Python 2.7 with support for following modules (those modules are commonly used and you probably don't need to do anything.)
  1. argparse
  2. subprocess
  3. re
  4. time
  5. os
  6. sys
• ffmpeg: Everyone should know what ffmpeg is. Downlolad static build for your OS from here.
• TAppEncoder: HM reference encoder. Download source code from here.
• TApp360Convert: Projection format conversion tool for 360 video. Download source code from here. This tool depends on HM software, therefore follow the descriptions in readme.txt to compile TApp360ConvertStatic together with TAppEncoderStatic.
• hevc2omaf: Command line tool for creation of OMAF content.

Compiled static binaries for TAppEncoder, TApp360Convert, and hevc2omaf should be located in ./bin/[linux|osx|win]

Please consider building these tools if they are missing in ./bin/[YourOS]

Also make sure that ffmpeg is installed on your system and is set in your $PATH

Description of the script

This script takes as input an ERP projected YUV file and creates the OMAF test vectors in the following 5 steps:

Step 1: convert ERP yuv to high res CMP yuv

    WARNING: This file is a raw video file with 6k resolution, it will consume a lot of storage.

Step 2: downlscale highres CMP yuv to additional lowres CMP yuv

Step 3: split both high and low res files into 24 tiles (each). This creates all required yuv tiles

Step 4: encode each tile as MCTS for provided QPs

    WARNING: if you use HM this process might consume a lot of time since HM reference software is not optimized for speed. Using HM you can also only encode a single QP. Consider using another encoder if you want to save some time or use multiple QPs.

Step 5: package encoded HEVC bitstreams to OMAF files

all the output is written to the output directory specified using -o command line argument. Use --help for more information on other arguments.

Example usage

Use kvazaar encoder and process 1017 frames of raw_video.yuv with two QPs 32 and 25 and use max 12 threads:

./create_omaf_files.py -s 1-5 -i raw_video.yuv -f 1017 -fr 30 -q 32 25 -t 12 --codec 1 -o kvazaarEncoding

Use HM encoder and only encode and package 270 frames of yuv files from 'folder/with/yuvs' directory and name the sequenze 'Garage'

./create_omaf_files.py -s 4-5 -i folder/with/yuvs -f 270 -fr 30 -q 32 -t 8 --codec 0 -c conf/encoder_randomaccess_main_RAP9.cfg -p Garage -o HMencodings

License

Please see LICENSE.txt file for the terms of use of the contents of this repository.

For more information, please contact: [email protected]

Copyright (c) 1995-2019 Fraunhofer-Gesellschaft zur Förderung der angewandten Forschung e.V.

All rights reserved.