Video compression is a challenge for a wide range of applications, as it reduces overall image quality and adds unwanted artefacts such as discoloration, break-ups and noise to the image. But when do you need uncompressed? And is compressed good enough?
In several high-end installations and projects there is a requirement for the media server to play back video uncompressed. However, in the majority of installations and projects, there is no need for uncompressed playback as it adds costs for disk space and I/O system. Smart use of compression yields virtually lossless playback (a well-trained eye can see the difference).
Why compress?
An uncompressed video file requires a lot from the media server, both as regards physical storage space as well as bandwidth (data transfer from hard-drive to graphics card or CPU) for video playback. Compressing the video file will reduce the file size as well as bandwidth requirement.
What is compression?
Video compression is based on the concept of codecs. A codec (which stands for compressor/decompressor) is the part of the editing and display technology responsible for storing and playing back compressed digitized video.
There’s a wide range of codecs available, each optimized for a particular kind of source material and playback requirements.
Image sequences
As an alternative to video playback, WATCHOUT can play image sequences. This is conceptually similar to video file playback, but instead of storing the entire video in a single file, each video frame is stored as a separate file.
For image sequences, WATCHOUT supports a number image file formats, such as BMP, TIFF, PNG, JPEG, DDS and Targa.
How do you play uncompressed video with WATCHOUT?
Due to the large amount of raw data in an uncompressed video stream, storing and playing back the video is usually not feasible. To play back uncompressed video from WATCHOUT we recommend the use of image sequences (TIFF for optimal playback performance).
Compression example
As an example of file size and image quality, we use the film Big Buck Bunny (downloaded from https://media.xiph.org).
The codecs used are MPEG-2 (4:2:0), H.264 4:2:0, HAP, HAPQ and of course uncompressed.
Image quality metrics
In addition to the differences in file size, we have an example of an image quality metric, where we use Peak-Signal-Noise-Ratio (PSNR). This is where we compare a single frame from the image sequence, and compare the noise in the various compressed formats.
The quality of the compressed files is close to identical and the difference will mainly be file size and playback performance.
|
|
Bit Rate (Mb/s) |
Disc Bandwidth (MB/s) |
Size per minute (MB) |
RMSE Error (%) |
|
H264 |
12,0 |
1,5 |
90,0 |
0,76% |
|
MPEG-2 |
25,0 |
3,1 |
187,5 |
0,87% |
|
HAP |
166,1 |
20,8 |
1245,8 |
0,68% |
|
HAPQ |
332,2 |
41,5 |
2491,7 |
0,52% |
|
Uncomp. |
1423,8 |
178,0 |
10678,7 |
0,00% |
As you can see the error rate is small, but a trained eye will see the difference. In HAP_Q the experienced image quality is higher – most of the measured errors are from YCoCg colorspace where colors are subject to errors.
