This movie is known to be notoriously hard to encode and therefore ideally fits my needs. I chose 3 scenes here: The beach landing (frames 8430 - 22281), a still scene in the controller's office (43790 - 45189) that has a lot of film grain and a night scene (frames 94122 - 106098). This should prove sufficiently difficult to get any codec into trouble. Let's see if I succeeded.
I don't think there's too much to be said about the beach scene. No steadycam, underwater shots, a lot of debris and confusion on the screen. Before the shooting starts there are many close-ups in the boat. As the boat door opens we can notice an immediate drop in details even in the source, so clearly this was a challenge to get this encoded for the DVD as well.
As we see the GIs approach the beach, the constant firing makes the scene look like it's raining. And as we've seen in Matrix3, no codec likes rain.
As we get on shore things start to shape up again. I decided to concentrate on the facial details:
Our final scene takes place at night, and black has proven to be notoriously hard to encode in the past.
Finally, our soldiers moving out during the night:
Conclusion
The results in SPR were remarkably similar to those of Matrix3. The NeroDigital codecs dominated, though not by as much as in Matrix3. They are followed by XviD and VP6.
I rather liked the newcomer ASP codec HDX4 until the night scene. It reminds me of early QPel implementations though. I recall disliking QPel quite a bit, but I suppose later implementations improved and made it more visually pleasing, and perhaps that's what we'll see about film grain.
3ivX and DivX once again performed similarly, and RV10 delivered the usual performance: no blocks, but a definite lack of details the farther you go from the camera.
WMV9 exhibited blocking and smearing that reminded me of DivX3. Plus there's the color bleeding. Finally, VSS delivered a varying performance, at times good, at times pretty bad. I think that it's just the rate control.
Now you can proceed to the Futurama test or go directly to the conclusion.
This document was last updated on December 29, 2004