引用:无损音频编码器性能对比

Performance comparison of lossless audio compressors

Updated: 2005-02-07 (YYYY-MM-DD)

On this page I want to compare the performance of lossless audio compressors available for Windows. These tests are limited to measuring the compression ratio and speed. There are other things that might influence the choice for a certain compressor, such as: availability for your platform / cross platform usability, open/closed source, error resistancy, nice GUI, on-the-fly ripping/compressing, random access while playing, low CPU usage while playing, hardware player for the format, etc. I'm not going into those differences here. You can find more information about the features of each compressor in this thread on Hydrogen Audio or on their websites:

The tests

How did I do the tests? I ripped 10 entire CDs with EAC's "CD-Image and cue-sheet" option. Then the CD image was compressed with one of the compressors. During this time I didn't run any other programs on the PC because this would influence the compression time. The compression factor varies widely with the kind of music that is compressed. Therefore I tried to test with various types of music. But my CD collection consists of mainly pop music, so my possibilities are limited (Guruboolez has done a similar test with classical music).

The Shorten tests are done with the ICL7.1 version compiled by John Edwards. It's much faster than the versions that are compiled under Cygwin.

Only 3 albums were tested with RKAU -l3 (maximum compression). It is so slow that I got bored testing it. But from this 3 albums you can get an idea of how much extra compression it has to offer over -l1 (fast compression).

Encoding with WMA 9 was done with this command line: cscript.exe wmcmd.vbs -input file.wav -output file.wma - a_codec WMA9LSL -a_mode 2 -a_setting Q100_44_2_16. Decoding was done with Foobar2000.

The results

The detailed test results are on a separate page. The overall results follow here:

Compressor Options Ratio* Encoding
Speed**
Decoding
Speed**
    % x realtime x realtime
La 0.4b default 55.5 2.1 2.7
OptimFROG 4.509 highnew 55.8 1.1 1.5
Monkey's Audio 3.99 extra high 56.4 8.8 8.7
OptimFROG 4.509 default 56.7 6.0 8.9
Monkey's Audio 3.99 high 56.9 15.8 14.3
Monkey's Audio 3.99 normal 57.3 18.1 16.0
WavPack 4.0 high 58.0 16.0 16.1
WMA 9 default 58.0 9.4 10.8
RKAU 1.07 fast (l1) 58.4 8.0 9.6
LPAC Archiver 1.41 medium, JS, random access 58.8 13.5 24.6
LPAC Archiver 1.41 extra high, JS, random access 58.8 8.1 20.2
TTA 3.2 default 59.0 26.6 23.4
WavPack 4.0 normal 59.4 26.3 28.4
FLAC 1.1.2 8 59.6 4.2 44.7
FLAC 1.1.2 default (5) 59.8 19.9 44.7
Apple Lossless (iTunes 4.7) automatic 60.0 15.6 35.0
Shorten 3.6.0 default 63.7 33.7 70.9

*Compression ratio = filesize of the compressed files / filesize of the wave files * 100%
**Speed = duration of an album / encoding c.q. decoding time. The speed is measured on my PC with an AMD Athlon 800 processor.

Best compression:

  1. La 0.4b (default)
  2. OptimFROG 4.509 (highnew)
  3. Monkey's Audio 3.99 (extra high)

Fastest encoding:

  1. Shorten 3.6.0
  2. TTA 3.2
  3. WavPack 4.0 (normal)

Fastest decoding:

  1. Shorten 3.6.0
  2. FLAC 1.1.2 (any compression level)
  3. Apple Lossless (iTunes 4.7)

Some remarks about the results:

  • La 0.4b offers the best compression followed shortly by OptimFROG at 'highnew' level. But the complicated algorithms used for such extreme compression ratios lead to slow encoding and decoding and a high CPU load during playback. Monkey's Audio at Extra High level is on the third place and OptimFROG default is fourth. Both are easy on the CPU during playback.
  • With LPAC and FLAC it doesn't really pay off to use the "Extra high" c.q. "-8" setting. It's much slower than "Medium" c.q. "-5", but the extra compression is little (to none in the case of LPAC with "random access" enabled).
  • Shorten, though popular in certain trading communities, offers poor compression.

Links

原文地址:https://www.cnblogs.com/gaozehua/p/2307575.html