•• Locking Exposure Prior to Entering LIVEVIEW vs Locking Exposure during LIVEVIEW••

[Robert Altman]

please don't quote me unless you're really quoting me. i never said such a thing, nor will i ever. quite the contrary. what i did say is that the trick doesn't necessarily make the video look better, for various reasons. i mean all those extra bits are used to encode the extra complexity, not for the "rest" of the image.

/matt

I am sorry for not quoting you, but the gist of what you have just said is that the actual compression quality is a constant, and that as you see higher bit rates it represents higher complexity scenes that will still have the same degree of compression artifacts.

Am I understanding your explanation?

If in fact it is highly detailed scenes that stress the algorithm the d90 uses perhaps using a mild soft effect filter in front of the lens could improve the image quality? I don't have any of these filters--has anyone tried them?

 

[mattsand]

I am sorry for not quoting you, but the gist of what you have just said is that the actual compression quality is a constant

yes, and that's pretty much all i said.

and that as you see higher bit rates it represents higher complexity scenes that will still have the same degree of compression artifacts

i don't know, sort of, but it depends on the scene. just play with your jpeg export settings in photoshop or whatever, you'll see that different quality settings give the same visual quality depending on the scene, and it's not necessarily based on what you think looks complex. remember, jpeg is based on removing or reducing the s/n ratio of frequencies we're not supposed to see anyway, but it's mathematical, there's no concious eye in the camera making the distinction. i took a lot of footage of sand dunes for my feature, and while that looks like a blur to the camera it's actually a lot of smooth gradients overlapping each other, a very complex signal, but the camera constantly compressed it as if it was a white wall. using the low contrast curve and pushing back in post helped a lot, since the blocks in the shadow and highlights areas are significantly reduced that way.

/matt

 

[mattsand]

That sounds like a 5D MKII endorsement.

h.264 has the same problems as avchd, which is a form of h.264, and it's quite obvious in 5d footage. it makes assumptions about what's image data and what's not and thus noise and texture is removed or reduced. compress some noisy d90 footage as h.264 and you'll see how much more like the 5d it starts to look. mpeg-2 isn't as efficient and has more artifacts at the same bitrate, but it produces images where texture is preserved, to me they look visually more like the original, and they don't "band" as much when you push the contrast.

/matt

 

[squig]

OK lets say the D90 had the option to select the current mjpeg codec or 17Mbps AVCHD which would choose for your film?

 

[mattsand]

avchd most likely. but if i had the option of 30mbit mjpeg i'd pick that.

/matt