Putting the Squeeze on LSX-MPEG Suite 2.0
The real test came when we passed raw video through the encoder — in this case, a video source file 1.1GB in size with 320x240 frames at 30fps. Unlike our digitally generated animation clip, which has the pristine signal-to-noise ratio MPEG likes to work with, raw footage takes more codec ingenuity in order to come across in the real world. We again chose our Windows 2000, Pentium III 500MHz PC as the test bed, nixing our original plan to use our trusty Windows NT PC. Although the Ligos player has been known to work with Windows NT, Ligos advises against it because there are some incompatibility issues with NT and DirectX. Die-hard NT users might want to try it, but make sure you have the latest version of NT Service Pack before proceeding.
We set up the encoder to compress at a 54:1 ratio to get the video to stream at 1Mbps (nearly Video CD quality), clicked on MPEG-1, and started the coding. The result? A majority of the near 10-minute video played back at decent quality. It scored a 48 on average image quality, and the speed of compression was better than expected. It took 28 minutes to compress, less than three times the original run time. Not bad, considering the encoder drew processing from the Pentium III and didn’t have an additional processing card to take some of the brunt of transcoding.
On another pass through the encoder, we decided to slow down the mode of motion estimation, a nifty option that can net higher quality video but takes more time to process. However, while it took almost twice as long to transcode for 1Mbps than with the previous test, we didn’t get twice the quality for the investment. The average image quality inched up only a notch, from 48 to 49. Hardly worth the wait.
Next, we ran the file through the Ligos encoder at a 108:1 compression ratio for 500Kbps, the high end of DSL exchange. A warning screen popped up, informing us that quality may be poor. We ran it anyway and found the playback to be usable in some cases, although the discerning eye could see that on high-motion segments, some of the solids in the background tended to appear "blocky," an effect common of macroblock-based, motion compensated Discrete Cosine Transform (DCT) coding. This one scored an average image quality of 31.
Hacker’s Dream/Novice’s Nightmare
Overall, we found this suite to be a codec hacker’s dream-come-true, but we wondered if the novice user might find the interface to be more of a nightmare. We could toy with motion vectors, change bit rates, and tweak to our heart’s content. We could pull up the profile manager feature as a starting point for settings, and even save customized settings and process in batches, yet some of the more basic functions proved to be difficult to perform.
Especially frustrating was the pop-up dialog box that appeared every time we attempted to change a number in the parameter boxes. It asks for a new "integer." Duh. When we wanted to enter a new, uh, integer, we had to first move the mouse to the pop-up box, click on the "OK" button, and then return to wherever we were in the parameters menu in order to put in a new number. Another frustration was that we couldn’t just select a specific format, like Video CD, from the main menu of options and expect the encoder to change all settings relative to the format. Instead, it required us to know what settings are needed for the format, and it let us get pretty far into the trial-and-error process before it prompted us that a certain resolution won’t work with a certain format, for example.
And while it is great that the encoder has a lot of choices for sequencing B, P and I frames and other advanced features, there is room for improvement on the support end. It would be nice to click on HELP at the Sequence window, and have it come back with a discussion on the sequencing of frames, instead of taking us all the way back to the beginning of the tutorial for us to find the answer.